/* * Support for Intel Camera Imaging ISP subsystem. * Copyright (c) 2015, Intel Corporation. * * This program is free software; you can redistribute it and/or modify it * under the terms and conditions of the GNU General Public License, * version 2, as published by the Free Software Foundation. * * This program is distributed in the hope it will be useful, but WITHOUT * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for * more details. */ #ifndef _REF_VECTOR_FUNC_H_INCLUDED_ #define _REF_VECTOR_FUNC_H_INCLUDED_ #ifdef INLINE_VECTOR_FUNC #define STORAGE_CLASS_REF_VECTOR_FUNC_H static inline #define STORAGE_CLASS_REF_VECTOR_DATA_H static inline_DATA #else /* INLINE_VECTOR_FUNC */ #define STORAGE_CLASS_REF_VECTOR_FUNC_H extern #define STORAGE_CLASS_REF_VECTOR_DATA_H extern_DATA #endif /* INLINE_VECTOR_FUNC */ #include "ref_vector_func_types.h" /* @brief Doubling multiply accumulate with saturation * * @param[in] acc accumulator * @param[in] a multiply input * @param[in] b multiply input * * @return acc + (a*b) * * This function will do a doubling multiply ont * inputs a and b, and will add the result to acc. * in case of an overflow of acc, it will saturate. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector2w OP_1w_maccd_sat( tvector2w acc, tvector1w a, tvector1w b ); /* @brief Doubling multiply accumulate * * @param[in] acc accumulator * @param[in] a multiply input * @param[in] b multiply input * * @return acc + (a*b) * * This function will do a doubling multiply ont * inputs a and b, and will add the result to acc. * in case of overflow it will not saturate but wrap around. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector2w OP_1w_maccd( tvector2w acc, tvector1w a, tvector1w b ); /* @brief Re-aligning multiply * * @param[in] a multiply input * @param[in] b multiply input * @param[in] shift shift amount * * @return (a*b)>>shift * * This function will multiply a with b, followed by a right * shift with rounding. the result is saturated and casted * to single precision. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w OP_1w_mul_realigning( tvector1w a, tvector1w b, tscalar1w shift ); /* @brief Leading bit index * * @param[in] a input * * @return index of the leading bit of each element * * This function finds the index of leading one (set) bit of the * input. The index starts with 0 for the LSB and can go upto * ISP_VEC_ELEMBITS-1 for the MSB. For an input equal to zero, * the returned index is -1. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w OP_1w_lod( tvector1w a); /* @brief Config Unit Input Processing * * @param[in] a input * @param[in] input_scale input scaling factor * @param[in] input_offset input offset factor * * @return scaled & offset added input clamped to MAXVALUE * * As part of input processing for piecewise linear estimation config unit, * this function will perform scaling followed by adding offset and * then clamping to the MAX InputValue * It asserts -MAX_SHIFT_1W <= input_scale <= MAX_SHIFT_1W, and * -MAX_SHIFT_1W <= input_offset <= MAX_SHIFT_1W */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w OP_1w_input_scaling_offset_clamping( tvector1w a, tscalar1w_5bit_signed input_scale, tscalar1w_5bit_signed input_offset); /* @brief Config Unit Output Processing * * @param[in] a output * @param[in] output_scale output scaling factor * * @return scaled & clamped output value * * As part of output processing for piecewise linear estimation config unit, * This function will perform scaling and then clamping to output * MAX value. * It asserts -MAX_SHIFT_1W <= output_scale <= MAX_SHIFT_1W */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w OP_1w_output_scaling_clamping( tvector1w a, tscalar1w_5bit_signed output_scale); /* @brief Config Unit Piecewiselinear estimation * * @param[in] a input * @param[in] config_points config parameter structure * * @return piecewise linear estimated output * * Given a set of N points {(x1,y1),()x2,y2), ....,(xn,yn)}, to find * the functional value at an arbitrary point around the input set, * this function will perform input processing followed by piecewise * linear estimation and then output processing to yield the final value. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w OP_1w_piecewise_estimation( tvector1w a, ref_config_points config_points); /* @brief Fast Config Unit * * @param[in] x input * @param[in] init_vectors LUT data structure * * @return piecewise linear estimated output * This block gets an input x and a set of input configuration points stored in a look-up * table of 32 elements. First, the x input is clipped to be within the range [x1, xn+1]. * Then, it computes the interval in which the input lies. Finally, the output is computed * by performing linear interpolation based on the interval properties (i.e. x_prev, slope, * and offset). This block assumes that the points are equally spaced and that the interval * size is a power of 2. **/ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w OP_1w_XCU( tvector1w x, xcu_ref_init_vectors init_vectors); /* @brief LXCU * * @param[in] x input * @param[in] init_vectors LUT data structure * * @return logarithmic piecewise linear estimated output. * This block gets an input x and a set of input configuration points stored in a look-up * table of 32 elements. It computes the interval in which the input lies. * Then output is computed by performing linear interpolation based on the interval * properties (i.e. x_prev, slope, * and offset). * This BBB assumes spacing x-coordinates of "init vectors" increase exponentially as * shown below. * interval size : 2^0 2^1 2^2 2^3 * x-coordinates: x0<--->x1<---->x2<---->x3<----> **/ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w OP_1w_LXCU( tvector1w x, xcu_ref_init_vectors init_vectors); /* @brief Coring * * @param[in] coring_vec Amount of coring based on brightness level * @param[in] filt_input Vector of input pixels on which Coring is applied * @param[in] m_CnrCoring0 Coring Level0 * * @return vector of filtered pixels after coring is applied * * This function will perform adaptive coring based on brightness level to * remove noise */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w coring( tvector1w coring_vec, tvector1w filt_input, tscalar1w m_CnrCoring0 ); /* @brief Normalised FIR with coefficients [3,4,1] * * @param[in] m 1x3 matrix with pixels * * @return filtered output * * This function will calculate the * Normalised FIR with coefficients [3,4,1], *-5dB at Fs/2, -90 degree phase shift (quarter pixel) */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x3m_5dB_m90_nrm ( const s_1w_1x3_matrix m); /* @brief Normalised FIR with coefficients [1,4,3] * * @param[in] m 1x3 matrix with pixels * * @return filtered output * * This function will calculate the * Normalised FIR with coefficients [1,4,3], *-5dB at Fs/2, +90 degree phase shift (quarter pixel) */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x3m_5dB_p90_nrm ( const s_1w_1x3_matrix m); /* @brief Normalised FIR with coefficients [1,2,1] * * @param[in] m 1x3 matrix with pixels * * @return filtered output * * This function will calculate the * Normalised FIR with coefficients [1,2,1], -6dB at Fs/2 */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x3m_6dB_nrm ( const s_1w_1x3_matrix m); /* @brief Normalised FIR with coefficients [13,16,3] * * @param[in] m 1x3 matrix with pixels * * @return filtered output * * This function will calculate the * Normalised FIR with coefficients [13,16,3], */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x3m_6dB_nrm_ph0 ( const s_1w_1x3_matrix m); /* @brief Normalised FIR with coefficients [9,16,7] * * @param[in] m 1x3 matrix with pixels * * @return filtered output * * This function will calculate the * Normalised FIR with coefficients [9,16,7], */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x3m_6dB_nrm_ph1 ( const s_1w_1x3_matrix m); /* @brief Normalised FIR with coefficients [5,16,11] * * @param[in] m 1x3 matrix with pixels * * @return filtered output * * This function will calculate the * Normalised FIR with coefficients [5,16,11], */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x3m_6dB_nrm_ph2 ( const s_1w_1x3_matrix m); /* @brief Normalised FIR with coefficients [1,16,15] * * @param[in] m 1x3 matrix with pixels * * @return filtered output * * This function will calculate the * Normalised FIR with coefficients [1,16,15], */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x3m_6dB_nrm_ph3 ( const s_1w_1x3_matrix m); /* @brief Normalised FIR with programable phase shift * * @param[in] m 1x3 matrix with pixels * @param[in] coeff phase shift * * @return filtered output * * This function will calculate the * Normalised FIR with coefficients [8-coeff,16,8+coeff], */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x3m_6dB_nrm_calc_coeff ( const s_1w_1x3_matrix m, tscalar1w_3bit coeff); /* @brief 3 tap FIR with coefficients [1,1,1] * * @param[in] m 1x3 matrix with pixels * * @return filtered output * * This function will calculate the * FIR with coefficients [1,1,1], -9dB at Fs/2 normalized with factor 1/2 */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x3m_9dB_nrm ( const s_1w_1x3_matrix m); #ifdef ISP2401 /* @brief symmetric 3 tap FIR acts as LPF or BSF * * @param[in] m 1x3 matrix with pixels * @param[in] k filter coefficient shift * @param[in] bsf_flag 1 for BSF and 0 for LPF * * @return filtered output * * This function performs variable coefficient symmetric 3 tap filter which can * be either used as Low Pass Filter or Band Stop Filter. * Symmetric 3tap tap filter with DC gain 1 has filter coefficients [a, 1-2a, a] * For LPF 'a' can be approximated as (1 - 2^(-k))/4, k = 0, 1, 2, ... * and filter output can be approximated as: * out_LPF = ((v00 + v02) - ((v00 + v02) >> k) + (2 * (v01 + (v01 >> k)))) >> 2 * For BSF 'a' can be approximated as (1 + 2^(-k))/4, k = 0, 1, 2, ... * and filter output can be approximated as: * out_BSF = ((v00 + v02) + ((v00 + v02) >> k) + (2 * (v01 - (v01 >> k)))) >> 2 * For a given filter coefficient shift 'k' and bsf_flag this function * behaves either as LPF or BSF. * All computation is done using 1w arithmetic and implementation does not use * any multiplication. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w sym_fir1x3m_lpf_bsf(s_1w_1x3_matrix m, tscalar1w k, tscalar_bool bsf_flag); #endif /* @brief Normalised 2D FIR with coefficients [1;2;1] * [1,2,1] * * @param[in] m 3x3 matrix with pixels * * @return filtered output * * This function will calculate the * Normalised FIR with coefficients [1;2;1] * [1,2,1] * Unity gain filter through repeated scaling and rounding * - 6 rotate operations per output * - 8 vector operations per output * _______ * 14 total operations */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir3x3m_6dB_nrm ( const s_1w_3x3_matrix m); /* @brief Normalised 2D FIR with coefficients [1;1;1] * [1,1,1] * * @param[in] m 3x3 matrix with pixels * * @return filtered output * * This function will calculate the * Normalised FIR with coefficients [1;1;1] * [1,1,1] * * (near) Unity gain filter through repeated scaling and rounding * - 6 rotate operations per output * - 8 vector operations per output * _______ * 14 operations */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir3x3m_9dB_nrm ( const s_1w_3x3_matrix m); /* @brief Normalised dual output 2D FIR with coefficients [1;2;1] * [1,2,1] * * @param[in] m 4x3 matrix with pixels * * @return two filtered outputs (2x1 matrix) * * This function will calculate the * Normalised FIR with coefficients [1;2;1] * [1,2,1] * and produce two outputs (vertical) * Unity gain filter through repeated scaling and rounding * compute two outputs per call to re-use common intermediates * - 4 rotate operations per output * - 6 vector operations per output (alternative possible, but in this * form it's not obvious to re-use variables) * _______ * 10 total operations */ STORAGE_CLASS_REF_VECTOR_FUNC_H s_1w_2x1_matrix fir3x3m_6dB_out2x1_nrm ( const s_1w_4x3_matrix m); /* @brief Normalised dual output 2D FIR with coefficients [1;1;1] * [1,1,1] * * @param[in] m 4x3 matrix with pixels * * @return two filtered outputs (2x1 matrix) * * This function will calculate the * Normalised FIR with coefficients [1;1;1] * [1,1,1] * and produce two outputs (vertical) * (near) Unity gain filter through repeated scaling and rounding * compute two outputs per call to re-use common intermediates * - 4 rotate operations per output * - 7 vector operations per output (alternative possible, but in this * form it's not obvious to re-use variables) * _______ * 11 total operations */ STORAGE_CLASS_REF_VECTOR_FUNC_H s_1w_2x1_matrix fir3x3m_9dB_out2x1_nrm ( const s_1w_4x3_matrix m); /* @brief Normalised 2D FIR 5x5 * * @param[in] m 5x5 matrix with pixels * * @return filtered output * * This function will calculate the * Normalised FIR with coefficients [1;1;1] * [1;2;1] * [1,2,1] * [1,1,1] * and produce a filtered output * (near) Unity gain filter through repeated scaling and rounding * - 20 rotate operations per output * - 28 vector operations per output * _______ * 48 total operations */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir5x5m_15dB_nrm ( const s_1w_5x5_matrix m); /* @brief Normalised FIR 1x5 * * @param[in] m 1x5 matrix with pixels * * @return filtered output * * This function will calculate the * Normalised FIR with coefficients [1,2,1] * [1,1,1] = [1,4,6,4,1] * and produce a filtered output * (near) Unity gain filter through repeated scaling and rounding * - 4 rotate operations per output * - 5 vector operations per output * _______ * 9 total operations */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x5m_12dB_nrm ( const s_1w_1x5_matrix m); /* @brief Normalised 2D FIR 5x5 * * @param[in] m 5x5 matrix with pixels * * @return filtered output * * This function will calculate the * Normalised FIR with coefficients [1;2;1] * [1;2;1] * [1,2,1] * [1,2,1] * and produce a filtered output * (near) Unity gain filter through repeated scaling and rounding * - 20 rotate operations per output * - 30 vector operations per output * _______ * 50 total operations */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir5x5m_12dB_nrm ( const s_1w_5x5_matrix m); /* @brief Approximate averaging FIR 1x5 * * @param[in] m 1x5 matrix with pixels * * @return filtered output * * This function will produce filtered output by * applying the filter coefficients (1/8) * [1,1,1,1,1] * _______ * 5 vector operations */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x5m_box ( s_1w_1x5_matrix m); /* @brief Approximate averaging FIR 1x9 * * @param[in] m 1x9 matrix with pixels * * @return filtered output * * This function will produce filtered output by * applying the filter coefficients (1/16) * [1,1,1,1,1,1,1,1,1] * _______ * 9 vector operations */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x9m_box ( s_1w_1x9_matrix m); /* @brief Approximate averaging FIR 1x11 * * @param[in] m 1x11 matrix with pixels * * @return filtered output * * This function will produce filtered output by * applying the filter coefficients (1/16) * [1,1,1,1,1,1,1,1,1,1,1] * _______ * 12 vector operations */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x11m_box ( s_1w_1x11_matrix m); /* @brief Symmetric 7 tap filter with normalization * * @param[in] in 1x7 matrix with pixels * @param[in] coeff 1x4 matrix with coefficients * @param[in] out_shift output pixel shift value for normalization * * @return symmetric 7 tap filter output * * This function performs symmetric 7 tap filter over input pixels. * Filter sum is normalized by shifting out_shift bits. * Filter sum: p0*c3 + p1*c2 + p2*c1 + p3*c0 + p4*c1 + p5*c2 + p6*c3 * is implemented as: (p0 + p6)*c3 + (p1 + p5)*c2 + (p2 + p4)*c1 + p3*c0 to * reduce multiplication. * Input pixels should to be scaled, otherwise overflow is possible during * addition */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x7m_sym_nrm(s_1w_1x7_matrix in, s_1w_1x4_matrix coeff, tvector1w out_shift); /* @brief Symmetric 7 tap filter with normalization at input side * * @param[in] in 1x7 matrix with pixels * @param[in] coeff 1x4 matrix with coefficients * * @return symmetric 7 tap filter output * * This function performs symmetric 7 tap filter over input pixels. * Filter sum: p0*c3 + p1*c2 + p2*c1 + p3*c0 + p4*c1 + p5*c2 + p6*c3 * = (p0 + p6)*c3 + (p1 + p5)*c2 + (p2 + p4)*c1 + p3*c0 * Input pixels and coefficients are in Qn format, where n = * ISP_VEC_ELEMBITS - 1 (ie Q15 for Broxton) * To avoid double precision arithmetic input pixel sum and final sum is * implemented using avgrnd and coefficient multiplication using qrmul. * Final result is in Qm format where m = ISP_VEC_ELEMBITS - 2 (ie Q14 for * Broxton) */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x7m_sym_innrm_approx(s_1w_1x7_matrix in, s_1w_1x4_matrix coeff); /* @brief Symmetric 7 tap filter with normalization at output side * * @param[in] in 1x7 matrix with pixels * @param[in] coeff 1x4 matrix with coefficients * * @return symmetric 7 tap filter output * * This function performs symmetric 7 tap filter over input pixels. * Filter sum: p0*c3 + p1*c2 + p2*c1 + p3*c0 + p4*c1 + p5*c2 + p6*c3 * = (p0 + p6)*c3 + (p1 + p5)*c2 + (p2 + p4)*c1 + p3*c0 * Input pixels are in Qn and coefficients are in Qm format, where n = * ISP_VEC_ELEMBITS - 2 and m = ISP_VEC_ELEMBITS - 1 (ie Q14 and Q15 * respectively for Broxton) * To avoid double precision arithmetic input pixel sum and final sum is * implemented using addsat and coefficient multiplication using qrmul. * Final sum is left shifted by 2 and saturated to produce result is Qm format * (ie Q15 for Broxton) */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x7m_sym_outnrm_approx(s_1w_1x7_matrix in, s_1w_1x4_matrix coeff); /* @brief 4 tap filter with normalization * * @param[in] in 1x4 matrix with pixels * @param[in] coeff 1x4 matrix with coefficients * @param[in] out_shift output pixel shift value for normalization * * @return 4 tap filter output * * This function performs 4 tap filter over input pixels. * Filter sum is normalized by shifting out_shift bits. * Filter sum: p0*c0 + p1*c1 + p2*c2 + p3*c3 */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x4m_nrm(s_1w_1x4_matrix in, s_1w_1x4_matrix coeff, tvector1w out_shift); /* @brief 4 tap filter with normalization for half pixel interpolation * * @param[in] in 1x4 matrix with pixels * * @return 4 tap filter output with filter tap [-1 9 9 -1]/16 * * This function performs 4 tap filter over input pixels. * Filter sum: -p0 + 9*p1 + 9*p2 - p3 * This filter implementation is completely free from multiplication and double * precision arithmetic. * Typical usage of this filter is to half pixel interpolation of Bezier * surface * */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x4m_bicubic_bezier_half(s_1w_1x4_matrix in); /* @brief 4 tap filter with normalization for quarter pixel interpolation * * @param[in] in 1x4 matrix with pixels * @param[in] coeff 1x4 matrix with coefficients * * @return 4 tap filter output * * This function performs 4 tap filter over input pixels. * Filter sum: p0*c0 + p1*c1 + p2*c2 + p3*c3 * To avoid double precision arithmetic we implemented multiplication using * qrmul and addition using avgrnd. Coefficients( c0 to c3) formats are assumed * to be: Qm, Qn, Qo, Qm, where m = n + 2 and o = n + 1. * Typical usage of this filter is to quarter pixel interpolation of Bezier * surface with filter coefficients:[-9 111 29 -3]/128. For which coefficient * values should be: [-9216/2^17 28416/2^15 1484/2^16 -3072/2^17] for * ISP_VEC_ELEMBITS = 16. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x4m_bicubic_bezier_quarter(s_1w_1x4_matrix in, s_1w_1x4_matrix coeff); /* @brief Symmetric 3 tap filter with normalization * * @param[in] in 1x3 matrix with pixels * @param[in] coeff 1x2 matrix with coefficients * @param[in] out_shift output pixel shift value for normalization * * @return symmetric 3 tap filter output * * This function performs symmetric 3 tap filter input pixels. * Filter sum is normalized by shifting out_shift bits. * Filter sum: p0*c1 + p1*c0 + p2*c1 * is implemented as: (p0 + p2)*c1 + p1*c0 to reduce multiplication. * Input pixels should to be scaled, otherwise overflow is possible during * addition */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x3m_sym_nrm(s_1w_1x3_matrix in, s_1w_1x2_matrix coeff, tvector1w out_shift); /* @brief Symmetric 3 tap filter with normalization * * @param[in] in 1x3 matrix with pixels * @param[in] coeff 1x2 matrix with coefficients * * @return symmetric 3 tap filter output * * This function performs symmetric 3 tap filter over input pixels. * Filter sum: p0*c1 + p1*c0 + p2*c1 = (p0 + p2)*c1 + p1*c0 * Input pixels are in Qn and coefficient c0 is in Qm and c1 is in Qn format, * where n = ISP_VEC_ELEMBITS - 1 and m = ISP_VEC_ELEMBITS - 2 ( ie Q15 and Q14 * respectively for Broxton) * To avoid double precision arithmetic input pixel sum is implemented using * avgrnd, coefficient multiplication using qrmul and final sum using addsat * Final sum is Qm format (ie Q14 for Broxton) */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w fir1x3m_sym_nrm_approx(s_1w_1x3_matrix in, s_1w_1x2_matrix coeff); /* @brief Mean of 1x3 matrix * * @param[in] m 1x3 matrix with pixels * * @return mean of 1x3 matrix * * This function calculates the mean of 1x3 pixels, * with a factor of 4/3. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w mean1x3m( s_1w_1x3_matrix m); /* @brief Mean of 3x3 matrix * * @param[in] m 3x3 matrix with pixels * * @return mean of 3x3 matrix * * This function calculates the mean of 3x3 pixels, * with a factor of 16/9. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w mean3x3m( s_1w_3x3_matrix m); /* @brief Mean of 1x4 matrix * * @param[in] m 1x4 matrix with pixels * * @return mean of 1x4 matrix * * This function calculates the mean of 1x4 pixels */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w mean1x4m( s_1w_1x4_matrix m); /* @brief Mean of 4x4 matrix * * @param[in] m 4x4 matrix with pixels * * @return mean of 4x4 matrix * * This function calculates the mean of 4x4 matrix with pixels */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w mean4x4m( s_1w_4x4_matrix m); /* @brief Mean of 2x3 matrix * * @param[in] m 2x3 matrix with pixels * * @return mean of 2x3 matrix * * This function calculates the mean of 2x3 matrix with pixels * with a factor of 8/6. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w mean2x3m( s_1w_2x3_matrix m); /* @brief Mean of 1x5 matrix * * @param[in] m 1x5 matrix with pixels * * @return mean of 1x5 matrix * * This function calculates the mean of 1x5 matrix with pixels * with a factor of 8/5. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w mean1x5m(s_1w_1x5_matrix m); /* @brief Mean of 1x6 matrix * * @param[in] m 1x6 matrix with pixels * * @return mean of 1x6 matrix * * This function calculates the mean of 1x6 matrix with pixels * with a factor of 8/6. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w mean1x6m( s_1w_1x6_matrix m); /* @brief Mean of 5x5 matrix * * @param[in] m 5x5 matrix with pixels * * @return mean of 5x5 matrix * * This function calculates the mean of 5x5 matrix with pixels * with a factor of 32/25. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w mean5x5m( s_1w_5x5_matrix m); /* @brief Mean of 6x6 matrix * * @param[in] m 6x6 matrix with pixels * * @return mean of 6x6 matrix * * This function calculates the mean of 6x6 matrix with pixels * with a factor of 64/36. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w mean6x6m( s_1w_6x6_matrix m); /* @brief Minimum of 4x4 matrix * * @param[in] m 4x4 matrix with pixels * * @return minimum of 4x4 matrix * * This function calculates the minimum of * 4x4 matrix with pixels. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w min4x4m( s_1w_4x4_matrix m); /* @brief Maximum of 4x4 matrix * * @param[in] m 4x4 matrix with pixels * * @return maximum of 4x4 matrix * * This function calculates the maximum of * 4x4 matrix with pixels. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w max4x4m( s_1w_4x4_matrix m); /* @brief SAD between two 3x3 matrices * * @param[in] a 3x3 matrix with pixels * * @param[in] b 3x3 matrix with pixels * * @return 3x3 matrix SAD * * This function calculates the sum of absolute difference between two matrices. * Both input pixels and SAD are normalized by a factor of SAD3x3_IN_SHIFT and * SAD3x3_OUT_SHIFT respectively. * Computed SAD is 1/(2 ^ (SAD3x3_IN_SHIFT + SAD3x3_OUT_SHIFT)) ie 1/16 factor * of original SAD and it's more precise than sad3x3m() */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w sad3x3m_precise( s_1w_3x3_matrix a, s_1w_3x3_matrix b); /* @brief SAD between two 3x3 matrices * * @param[in] a 3x3 matrix with pixels * * @param[in] b 3x3 matrix with pixels * * @return 3x3 matrix SAD * * This function calculates the sum of absolute difference between two matrices. * This version saves cycles by avoiding input normalization and wide vector * operation during sum computation * Input pixel differences are computed by absolute of rounded, halved * subtraction. Normalized sum is computed by rounded averages. * Computed SAD is (1/2)*(1/16) = 1/32 factor of original SAD. Factor 1/2 comes * from input halving operation and factor 1/16 comes from mean operation */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w sad3x3m( s_1w_3x3_matrix a, s_1w_3x3_matrix b); /* @brief SAD between two 5x5 matrices * * @param[in] a 5x5 matrix with pixels * * @param[in] b 5x5 matrix with pixels * * @return 5x5 matrix SAD * * Computed SAD is = 1/32 factor of original SAD. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w sad5x5m( s_1w_5x5_matrix a, s_1w_5x5_matrix b); /* @brief Absolute gradient between two sets of 1x5 matrices * * @param[in] m0 first set of 1x5 matrix with pixels * @param[in] m1 second set of 1x5 matrix with pixels * * @return absolute gradient between two 1x5 matrices * * This function computes mean of two input 1x5 matrices and returns * absolute difference between two mean values. */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w absgrad1x5m(s_1w_1x5_matrix m0, s_1w_1x5_matrix m1); /* @brief Bi-linear Interpolation optimized(approximate) * * @param[in] a input0 * @param[in] b input1 * @param[in] c cloned weight factor * * @return (a-b)*c + b * * This function will do bi-linear Interpolation on * inputs a and b using constant weight factor c * * Inputs a,b are assumed in S1.15 format * Weight factor has to be in range [0,1] and is assumed to be in S2.14 format * * The bilinear interpolation equation is (a*c) + b*(1-c), * But this is implemented as (a-b)*c + b for optimization */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w OP_1w_bilinear_interpol_approx_c( tvector1w a, tvector1w b, tscalar1w_weight c); /* @brief Bi-linear Interpolation optimized(approximate) * * @param[in] a input0 * @param[in] b input1 * @param[in] c weight factor * * @return (a-b)*c + b * * This function will do bi-linear Interpolation on * inputs a and b using weight factor c * * Inputs a,b are assumed in S1.15 format * Weight factor has to be in range [0,1] and is assumed to be in S2.14 format * * The bilinear interpolation equation is (a*c) + b*(1-c), * But this is implemented as (a-b)*c + b for optimization */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w OP_1w_bilinear_interpol_approx( tvector1w a, tvector1w b, tvector1w_weight c); /* @brief Bi-linear Interpolation * * @param[in] a input0 * @param[in] b input1 * @param[in] c weight factor * * @return (a*c) + b*(1-c) * * This function will do bi-linear Interpolation on * inputs a and b using weight factor c * * Inputs a,b are assumed in S1.15 format * Weight factor has to be in range [0,1] and is assumed to be in S2.14 format * * The bilinear interpolation equation is (a*c) + b*(1-c), */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w OP_1w_bilinear_interpol( tvector1w a, tvector1w b, tscalar1w_weight c); /* @brief Generic Block Matching Algorithm * @param[in] search_window pointer to input search window of 16x16 pixels * @param[in] ref_block pointer to input reference block of 8x8 pixels, where N<=M * @param[in] output pointer to output sads * @param[in] search_sz search size for SAD computation * @param[in] ref_sz block size * @param[in] pixel_shift pixel shift to search the data * @param[in] search_block_sz search window block size * @param[in] shift shift value, with which the output is shifted right * * @return 0 when the computation is successful. * * This function compares the reference block with a block of size NxN in the search * window. Sum of absolute differences for each pixel in the reference block and the * corresponding pixel in the search block. Whole search window os traversed with the * reference block with the given pixel shift. * */ STORAGE_CLASS_REF_VECTOR_FUNC_H int generic_block_matching_algorithm( tscalar1w **search_window, tscalar1w **ref_block, tscalar1w *output, int search_sz, int ref_sz, int pixel_shift, int search_block_sz, tscalar1w_4bit_bma_shift shift); #ifndef ISP2401 /* @brief OP_1w_asp_bma_16_1_32way #else /* @brief OP_1w_asp_bma_16_1_32way_nomask #endif * * @param[in] search_area input search window of 16x16 pixels * @param[in] input_block input reference block of 8x8 pixels, where N<=M * @param[in] shift shift value, with which the output is shifted right * * @return 81 SADs for all the search blocks. * This function compares the reference block with a block of size 8x8 pixels in the * search window of 16x16 pixels. Sum of absolute differences for each pixel in the * reference block and the corresponding pixel in the search block is calculated. * Whole search window is traversed with the reference block with the pixel shift of 1 * pixels. The output is right shifted with the given shift value. The shift value is * a 4 bit value. * */ #ifndef ISP2401 STORAGE_CLASS_REF_VECTOR_FUNC_H bma_output_16_1 OP_1w_asp_bma_16_1_32way( #else STORAGE_CLASS_REF_VECTOR_FUNC_H bma_output_16_1 OP_1w_asp_bma_16_1_32way_nomask( #endif bma_16x16_search_window search_area, ref_block_8x8 input_block, tscalar1w_4bit_bma_shift shift); #ifndef ISP2401 /* @brief OP_1w_asp_bma_16_2_32way #else /* @brief OP_1w_asp_bma_16_2_32way_nomask #endif * * @param[in] search_area input search window of 16x16 pixels * @param[in] input_block input reference block of 8x8 pixels, where N<=M * @param[in] shift shift value, with which the output is shifted right * * @return 25 SADs for all the search blocks. * This function compares the reference block with a block of size 8x8 in the search * window of 16x61. Sum of absolute differences for each pixel in the reference block * and the corresponding pixel in the search block is computed. Whole search window is * traversed with the reference block with the given pixel shift of 2 pixels. The output * is right shifted with the given shift value. The shift value is a 4 bit value. * */ #ifndef ISP2401 STORAGE_CLASS_REF_VECTOR_FUNC_H bma_output_16_2 OP_1w_asp_bma_16_2_32way( #else STORAGE_CLASS_REF_VECTOR_FUNC_H bma_output_16_2 OP_1w_asp_bma_16_2_32way_nomask( #endif bma_16x16_search_window search_area, ref_block_8x8 input_block, tscalar1w_4bit_bma_shift shift); #ifndef ISP2401 /* @brief OP_1w_asp_bma_14_1_32way #else /* @brief OP_1w_asp_bma_14_1_32way_nomask #endif * * @param[in] search_area input search block of 16x16 pixels with search window of 14x14 pixels * @param[in] input_block input reference block of 8x8 pixels, where N<=M * @param[in] shift shift value, with which the output is shifted right * * @return 49 SADs for all the search blocks. * This function compares the reference block with a block of size 8x8 in the search * window of 14x14. Sum of absolute differences for each pixel in the reference block * and the corresponding pixel in the search block. Whole search window is traversed * with the reference block with 2 pixel shift. The output is right shifted with the * given shift value. The shift value is a 4 bit value. Input is always a 16x16 block * but the search window is 14x14, with last 2 pixels of row and column are not used * for computation. * */ #ifndef ISP2401 STORAGE_CLASS_REF_VECTOR_FUNC_H bma_output_14_1 OP_1w_asp_bma_14_1_32way( #else STORAGE_CLASS_REF_VECTOR_FUNC_H bma_output_14_1 OP_1w_asp_bma_14_1_32way_nomask( #endif bma_16x16_search_window search_area, ref_block_8x8 input_block, tscalar1w_4bit_bma_shift shift); #ifndef ISP2401 /* @brief OP_1w_asp_bma_14_2_32way #else /* @brief OP_1w_asp_bma_14_2_32way_nomask #endif * * @param[in] search_area input search block of 16x16 pixels with search window of 14x14 pixels * @param[in] input_block input reference block of 8x8 pixels, where N<=M * @param[in] shift shift value, with which the output is shifted right * * @return 16 SADs for all the search blocks. * This function compares the reference block with a block of size 8x8 in the search * window of 14x14. Sum of absolute differences for each pixel in the reference block * and the corresponding pixel in the search block. Whole search window is traversed * with the reference block with 2 pixels shift. The output is right shifted with the * given shift value. The shift value is a 4 bit value. * */ #ifndef ISP2401 STORAGE_CLASS_REF_VECTOR_FUNC_H bma_output_14_2 OP_1w_asp_bma_14_2_32way( #else STORAGE_CLASS_REF_VECTOR_FUNC_H bma_output_14_2 OP_1w_asp_bma_14_2_32way_nomask( #endif bma_16x16_search_window search_area, ref_block_8x8 input_block, tscalar1w_4bit_bma_shift shift); #ifdef ISP2401 /* @brief multiplex addition and passing * * @param[in] _a first pixel * @param[in] _b second pixel * @param[in] _c condition flag * * @return (_a + _b) if condition flag is true * _a if condition flag is false * * This function does multiplex addition depending on the input condition flag */ STORAGE_CLASS_REF_VECTOR_FUNC_H tvector1w OP_1w_cond_add( tvector1w _a, tvector1w _b, tflags _c); #endif #ifdef HAS_bfa_unit /* @brief OP_1w_single_bfa_7x7 * * @param[in] weights - spatial and range weight lut * @param[in] threshold - threshold plane, for range weight scaling * @param[in] central_pix - central pixel plane * @param[in] src_plane - src pixel plane * * @return Bilateral filter output * * This function implements, 7x7 single bilateral filter. * Output = {sum(pixel * weight), sum(weight)} * Where sum is summation over 7x7 block set. * weight = spatial weight * range weight * spatial weights are loaded from spatial_weight_lut depending on src pixel * position in the 7x7 block * range weights are computed by table look up from range_weight_lut depending * on scaled absolute difference between src and central pixels. * threshold is used as scaling factor. range_weight_lut consists of * BFA_RW_LUT_SIZE numbers of LUT entries to model any distribution function. * Piecewise linear approximation technique is used to compute range weight * It computes absolute difference between central pixel and 61 src pixels. */ STORAGE_CLASS_REF_VECTOR_FUNC_H bfa_7x7_output OP_1w_single_bfa_7x7( bfa_weights weights, tvector1w threshold, tvector1w central_pix, s_1w_7x7_matrix src_plane); /* @brief OP_1w_joint_bfa_7x7 * * @param[in] weights - spatial and range weight lut * @param[in] threshold0 - 1st threshold plane, for range weight scaling * @param[in] central_pix0 - 1st central pixel plane * @param[in] src0_plane - 1st pixel plane * @param[in] threshold1 - 2nd threshold plane, for range weight scaling * @param[in] central_pix1 - 2nd central pixel plane * @param[in] src1_plane - 2nd pixel plane * * @return Joint bilateral filter output * * This function implements, 7x7 joint bilateral filter. * Output = {sum(pixel * weight), sum(weight)} * Where sum is summation over 7x7 block set. * weight = spatial weight * range weight * spatial weights are loaded from spatial_weight_lut depending on src pixel * position in the 7x7 block * range weights are computed by table look up from range_weight_lut depending * on sum of scaled absolute difference between central pixel and two src pixel * planes. threshold is used as scaling factor. range_weight_lut consists of * BFA_RW_LUT_SIZE numbers of LUT entries to model any distribution function. * Piecewise linear approximation technique is used to compute range weight * It computes absolute difference between central pixel and 61 src pixels. */ STORAGE_CLASS_REF_VECTOR_FUNC_H bfa_7x7_output OP_1w_joint_bfa_7x7( bfa_weights weights, tvector1w threshold0, tvector1w central_pix0, s_1w_7x7_matrix src0_plane, tvector1w threshold1, tvector1w central_pix1, s_1w_7x7_matrix src1_plane); /* @brief bbb_bfa_gen_spatial_weight_lut * * @param[in] in - 7x7 matrix of spatial weights * @param[in] out - generated LUT * * @return None * * This function implements, creates spatial weight look up table used * for bilaterl filter instruction. */ STORAGE_CLASS_REF_VECTOR_FUNC_H void bbb_bfa_gen_spatial_weight_lut( s_1w_7x7_matrix in, tvector1w out[BFA_MAX_KWAY]); /* @brief bbb_bfa_gen_range_weight_lut * * @param[in] in - input range weight, * @param[in] out - generated LUT * * @return None * * This function implements, creates range weight look up table used * for bilaterl filter instruction. * 8 unsigned 7b weights are represented in 7 16bits LUT * LUT formation is done as follows: * higher 8 bit: Point(N) = Point(N+1) - Point(N) * lower 8 bit: Point(N) = Point(N) * Weight function can be any monotonic decreasing function for x >= 0 */ STORAGE_CLASS_REF_VECTOR_FUNC_H void bbb_bfa_gen_range_weight_lut( tvector1w in[BFA_RW_LUT_SIZE+1], tvector1w out[BFA_RW_LUT_SIZE]); #endif #ifdef ISP2401 /* @brief OP_1w_imax32 * * @param[in] src - structure that holds an array of 32 elements. * * @return maximum element among input array. * *This function gets maximum element from an array of 32 elements. */ STORAGE_CLASS_REF_VECTOR_FUNC_H int OP_1w_imax32( imax32_ref_in_vector src); /* @brief OP_1w_imaxidx32 * * @param[in] src - structure that holds a vector of elements. * * @return index of first element with maximum value among array. * * This function gets index of first element with maximum value * from 32 elements. */ STORAGE_CLASS_REF_VECTOR_FUNC_H int OP_1w_imaxidx32( imax32_ref_in_vector src); #endif #ifndef INLINE_VECTOR_FUNC #define STORAGE_CLASS_REF_VECTOR_FUNC_C #define STORAGE_CLASS_REF_VECTOR_DATA_C const #else /* INLINE_VECTOR_FUNC */ #define STORAGE_CLASS_REF_VECTOR_FUNC_C STORAGE_CLASS_REF_VECTOR_FUNC_H #define STORAGE_CLASS_REF_VECTOR_DATA_C STORAGE_CLASS_REF_VECTOR_DATA_H #include "ref_vector_func.c" #define VECTOR_FUNC_INLINED #endif /* INLINE_VECTOR_FUNC */ #endif /*_REF_VECTOR_FUNC_H_INCLUDED_*/