/**
* @file			sharp.c
* @brief		sharpening algorithm
* @author		Patrick Roth - roth@stettbacher.ch
* @version		1.0
* @date			2015-08-27
* @copyright	2012-2016 Stettbacher Signal Processing AG
* 
* @remarks
*
* <PRE>
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
* </PRE>
*
*/

#include <stdio.h>
#include <string.h>

#if (WITH_SIMD == 1)
#include <immintrin.h>		// see /usr/lib64/gcc/x86_64-suse-linux/4.7/include/immintrin.h
#endif // WITH_SIMD

#include "color_pipe_private.h"
#include "color.h"
#include "filter.h"


/**
 * Transform monochrome to YUV image.
 * 
 * @param img_yuv On return: image in YUV color space
 * @param img_mono input monochrome image
 * @param height image height in number of pixels
 * @param width image width in number of pixels
 * @param bit_channel monochrome bit resolution
 */
static void mono_to_yuv(int16_t *img_yuv, const void *img_mono, const int height, const int width, const int bit_channel) {
	
	int y, x, index_mono, index_yuv;
	const uint8_t *in8 = img_mono;
	const uint16_t *in16 = img_mono;
	
	
	index_mono = 0;
	index_yuv = 0;
	for(y = 0; y < height; y++) {
		for(x = 0; x < width; x++) {
			if(bit_channel <= 8) {
				img_yuv[index_yuv] = in8[index_mono];
				img_yuv[index_yuv+1] = 0;
				img_yuv[index_yuv+2] = 0;
			}
			else if(bit_channel <= 16) {
				img_yuv[index_yuv] = in16[index_mono];
				img_yuv[index_yuv+1] = 0;
				img_yuv[index_yuv+2] = 0;
			}
			index_mono++;
			index_yuv += 3;
		}
	}
}


/**
 * Transform YUV to monochrome image.
 * 
 * @param img_mono On return: monochrome image
 * @param img_yuv input monochrome image
 * @param height image height in number of pixels
 * @param width image width in number of pixels
 * @param bit_channel monochrome bit resolution
 */
static void yuv_to_mono(void *img_mono, const int16_t *img_yuv, const int height, const int width, const int bit_channel) {
	int y, x, index_mono, index_yuv;
	int8_t *out8 = img_mono;
	int16_t *out16 = img_mono;
	const int pix_max = (1<<bit_channel)-1;
	int value;
	
	index_mono = 0;
	index_yuv = 0;
	
	for(y = 0; y < height; y++) {
		for(x = 0; x < width; x++) {
			value = img_yuv[index_yuv];
			
			if(value < 0) {
				value = 0;
			}
			else if(value > pix_max) {
				value = pix_max;
			}
			
			if(bit_channel <= 8) {
				out8[index_mono] = value;
			}
			else if(bit_channel <= 16) {
				out16[index_mono] = value;
			}
			
			index_mono++;
			index_yuv += 3;
		}
	}
}


/**
 * Sharp pixel by applying 3x3 filter kernel. Fixed-point is used.
 * The kernel weights around the center are equal.
 * 
 * @param a_other kernel weight around center
 * @param a_center kernel weight at center position
 * @param p11 pixel value at position 1/1
 * @param p12 pixel value at position 1/2
 * @param p13 pixel value at position 1/3
 * @param p21 pixel value at position 2/1
 * @param p22 pixel value at position 2/2
 * @param p23 pixel value at position 2/3
 * @param p31 pixel value at position 3/1
 * @param p32 pixel value at position 3/2
 * @param p33 pixel value at position 3/3 
 * @param shift_fact The shifting factor defines how many number of bits the kernel and pixel were shifted to left.
 * @return filtered pixel value
 */
static inline int16_t do_sharp(const int16_t a_other, const int16_t a_center,
							   const int16_t p11, const int16_t p12, const int16_t p13,
							   const int16_t p21, const int16_t p22, const int16_t p23,
							   const int16_t p31, const int16_t p32, const int16_t p33,
							   const int shift_fact) {
	
	int16_t out;
	
	
	out = (	a_other*p11 + a_other*p12 + a_other*p13 +
			a_other*p21 + a_center*p22 + a_other*p23 +
			a_other*p31 + a_other*p32 + a_other*p33) >> shift_fact;
	return out;
}

#if 0
/*
 * Sharpening algorithm by using SSE instructions.
 * It's slower than the scalar algorithm above!!
 */
static int16_t do_sharp_sse(__m128i coeff_line0, __m128i coeff_line1, __m128i coeff_line2,
							__m128i px_line0, __m128i px_line1, __m128i px_line2, __m128i mask,
							int shift_fact) {
	
	__m128i y_line0, y_line1, y_line2, madd_line0, madd_line1, madd_line2;
	int32_t res[2];
	
	
	y_line0 = _mm_shuffle_epi8(px_line0, mask);
	y_line1 = _mm_shuffle_epi8(px_line1, mask);
	y_line2 = _mm_shuffle_epi8(px_line2, mask);
	
	madd_line0 = _mm_madd_epi16(y_line0, coeff_line0);
	madd_line1 = _mm_madd_epi16(y_line1, coeff_line1);
	madd_line2 = _mm_madd_epi16(y_line2, coeff_line2);
	
	madd_line0 = _mm_hadd_epi32(madd_line0, madd_line1);
	madd_line0 = _mm_hadd_epi32(madd_line0, madd_line2);
	
	madd_line0 = _mm_hadd_epi32(madd_line0, madd_line0);
	madd_line0 = _mm_hadd_epi32(madd_line0, madd_line0);
	
	_mm_storel_epi64((__m128i*)&res, madd_line0);
	return (res[0] >> shift_fact);
}
#endif

/**
 * Make given YUV image sharper. Use the given the filter strength to tune the sharpening strength.
 * If the local sharpening is set only those pixels are sharpened defined at the sharpening mask.
 * 
 * This sharpening algorithm high-pass filters the input image and adds it to itself. Therfore all edges become sharper.
 * The sharpening is done on the Y-channel only. The brightness is of interest. The U and V channel won't be touched to avoid color
 * shiftings.
 * 
 * @param img_out On return: high-pass filtered YUV image
 * @param img_in YUV image to filter with given kernel
 * @param height image height in number of pixels
 * @param width image width in number of pixels
 * @param sharp_strength sharpening strength factor
 * @param max_y maximum Y-channel value (depends on bit per pixel)
 * @param local_flag not 0 if local sharpening must be done based on sharpening mask
 * @param sharp_mask sharpening mask (binary image)
 */
static void make_sharper(int16_t *img_out, const int16_t * img_in, const int height, const int width, const float sharp_strength, const int max_y,
						 const int local_flag, const int8_t *sharp_mask) {
	
	int y, x, index_upper, index_center, index_lower;
	int16_t filter_output;
// 	__m128i coeff_line0, coeff_line1, coeff_line2, mask;
	
	
	/*
	 * don't touch it or check high-pass filter coefficient a_center for overflow!!
	 */
	const int shift_fact = 10;
	
	/*
	 * High-pass filter coefficients
	 * 
	 * e. g. shift_fact = 10 and sharp_strength = 4
	 * 		--> a_other = -4/8.0 * 2^10 = 2^9 	--> no overflow
	 * 		--> a_center = 4*2^10 = 2^12		--> no overflow
	 * 
	 * e. g. shift_fact = 10 and sharp_strength = 32
	 * 		--> a_other = -32/8.0 * 2^10 = 2^12	--> no overflow
	 * 		--> a_center = 32*2^10 = 2^15		--> overflow because this value cis not possible with int16_t datatype
	 */
	const int16_t a_other = -1.0*sharp_strength/8.0*(1<<shift_fact);
	const int16_t a_center = sharp_strength*(1<<shift_fact);
	
	
	/*
	coeff_line0 = _mm_set_epi16(0, 0, 0, 0, 0, a_other, a_other, a_other);
	coeff_line1 = _mm_set_epi16(0, 0, 0, 0, 0, a_other, a_center, a_other);
	coeff_line2 = _mm_set_epi16(0, 0, 0, 0, 0, a_other, a_other, a_other);
	mask = _mm_set_epi8(-1, -1, -1, -1, -1, -1, -1, -1, -1, -1, 13, 12, 7, 6, 1, 0);
	*/
	
	for(y = 1; y < (height-1); y++) {
		
		index_upper = (y-1)*width*3;
		index_center = y*width*3;
		index_lower = (y+1)*width*3;
		
		
		for(x = 1; x < (width-1); x++) {
			
			if(local_flag != 0 && sharp_mask[index_center+3] == 0) {
				// don't sharp this pixel
				filter_output = 0;
			}
			else {
				filter_output = do_sharp(a_other, a_center,
										 img_in[index_upper], img_in[index_upper+3], img_in[index_upper+6],
										 img_in[index_center], img_in[index_center+3], img_in[index_center+6],
										 img_in[index_lower], img_in[index_lower+3], img_in[index_lower+6],
										 shift_fact);
				/*
				filter_output = do_sharp_sse(	coeff_line0, coeff_line1, coeff_line2,
												_mm_lddqu_si128((__m128i*)(&img_in[index_upper])),
												_mm_lddqu_si128((__m128i*)(&img_in[index_center])),
												_mm_lddqu_si128((__m128i*)(&img_in[index_lower])),
												mask,
												shift_fact);
				*/

			}
			
			filter_output += img_in[index_center+3];
			
			img_out[index_center+3] = filter_output;
			img_out[index_center+4] = img_in[index_center+4];
			img_out[index_center+5] = img_in[index_center+5];
			index_upper += 3;
			index_center += 3;
			index_lower += 3;
		}
	}
	
	// handle horizontal upper border line (without corners)
	index_center = 0;
	index_upper = width*3;
	index_lower = index_upper;
	for(x = 1; x < (width-1); x++) {
		if(local_flag != 0 && sharp_mask[index_center+3] == 0) {
			// don't sharp this pixel
			filter_output = 0;
		}
		else {
			filter_output = do_sharp(a_other, a_center,
									 img_in[index_upper], img_in[index_upper+3], img_in[index_upper+6],
									 img_in[index_center], img_in[index_center+3], img_in[index_center+6],
									 img_in[index_lower], img_in[index_lower+3], img_in[index_lower+6],
									 shift_fact);
		}
		
		filter_output += img_in[index_center+3];
		
		img_out[index_center+3] = filter_output;
		img_out[index_center+4] = img_in[index_center+4];
		img_out[index_center+5] = img_in[index_center+5];
		index_upper += 3;
		index_center += 3;
		index_lower += 3;
	}
	
	// handle horizontal lower border line (without corners)
	index_center = (height-1)*width*3;
	index_upper = (height-2)*width*3;
	index_lower = index_upper;
	for(x = 1; x < (width-1); x++) {
		if(local_flag != 0 && sharp_mask[index_center+3] == 0) {
			// don't sharp this pixel
			filter_output = 0;
		}
		else {
			filter_output = do_sharp(a_other, a_center,
									 img_in[index_upper], img_in[index_upper+3], img_in[index_upper+6],
									 img_in[index_center], img_in[index_center+3], img_in[index_center+6],
									 img_in[index_lower], img_in[index_lower+3], img_in[index_lower+6],
									 shift_fact);
		}
		
		filter_output += img_in[index_center+3];
		
		img_out[index_center+3] = filter_output;
		img_out[index_center+4] = img_in[index_center+4];
		img_out[index_center+5] = img_in[index_center+5];
		index_upper += 3;
		index_center += 3;
		index_lower += 3;
	}
	
	// handle vertical left border line (without corners)
	for(y = 1; y < (height-1); y++) {
		index_upper = (y-1)*width*3;
		index_center = y*width*3;
		index_lower = (y+1)*width*3;
		
		if(local_flag != 0 && sharp_mask[index_center+3] == 0) {
			// don't sharp this pixel
			filter_output = 0;
		}
		else {
			filter_output = do_sharp(a_other, a_center,
									 img_in[index_upper+3], img_in[index_upper], img_in[index_upper+3],
									 img_in[index_center+3], img_in[index_center], img_in[index_center+3],
									 img_in[index_lower+3], img_in[index_lower], img_in[index_lower+3],
									 shift_fact);
		}
		
		filter_output += img_in[index_center];
		img_out[index_center] = filter_output;
		img_out[index_center+1] = img_in[index_center+1];
		img_out[index_center+2] = img_in[index_center+2];
	}
	
	// handle vertical right border line (without corners)
	for(y = 1; y < (height-1); y++) {
		index_upper = y*width*3-3;
		index_center = (y+1)*width*3-3;
		index_lower = (y+2)*width*3-3;
		
		if(local_flag != 0 && sharp_mask[index_center+3] == 0) {
			// don't sharp this pixel
			filter_output = 0;
		}
		else {
			filter_output = do_sharp(a_other, a_center,
									 img_in[index_upper-3], img_in[index_upper], img_in[index_upper-3],
									 img_in[index_center-3], img_in[index_center], img_in[index_center-3],
									 img_in[index_lower-3], img_in[index_lower], img_in[index_lower-3],
									 shift_fact);
		}
			
		filter_output += img_in[index_center];
		img_out[index_center] = filter_output;
		img_out[index_center+1] = img_in[index_center+1];
		img_out[index_center+2] = img_in[index_center+2];
	}
	
	
	/*
	 * Image corners are not sharpened!!
	 */
	
	// handle upper left corner
	img_out[0] = img_in[0];
	img_out[1] = img_in[1];
	img_out[2] = img_in[2];
	
	// handle upper right corner
	index_center = width*3-3;
	img_out[index_center] = img_in[index_center];
	img_out[index_center+1] = img_in[index_center+1];
	img_out[index_center+2] = img_in[index_center+2];
	
	// handle lower left corner
	index_center = (height-1)*width*3;
	img_out[index_center] = img_in[index_center];
	img_out[index_center+1] = img_in[index_center+1];
	img_out[index_center+2] = img_in[index_center+2];
	
	// handle lower right corner
	index_center = height*width*3-3;
	img_out[index_center] = img_in[index_center];
	img_out[index_center+1] = img_in[index_center+1];
	img_out[index_center+2] = img_in[index_center+2];
}


/**
 * Sharpening algorithm.
 * 
 * @param sharp_data required sharpening data
 * @return 0 on success otherwise -1
 */
int sharpening(struct sharp_data_t *sharp_data) {
	
	void *img_sharp, *img_unsharp;
	int is_color, bit_channel, width, height;
	int16_t *img_yuv, *img_yuv_sharp, *img_sobel, *img_gauss;
	float sharp_factor;
	enum sharp_alg_t sharp_alg;
	int8_t *sharp_mask;
	int local_sens;
	
	
	// put variables on stack
	is_color = sharp_data->is_color;
	img_sharp = sharp_data->img_sharp;
	img_unsharp = sharp_data->img_in;
	bit_channel = sharp_data->bit_channel;
	width = sharp_data->width;
	height = sharp_data->height;
	img_yuv = sharp_data->img_yuv;
	sharp_factor = sharp_data->sharp_factor;
	sharp_alg = sharp_data->sharp_alg;
	local_sens = sharp_data->local_sens;
	img_yuv_sharp = sharp_data->img_yuv_sharp;
	img_sobel = sharp_data->img_sobel;
	img_gauss = sharp_data->img_gauss;
	sharp_mask = sharp_data->sharp_mask;
	
	
	/*
	 * Sharpening is done on Y-channel.
	 * In case of color image, the RGB is transformed to YUV. In case of monochrom image,
	 * the Y-channel is used only.
	 */
	if(is_color) {
		// RGB to YUV transformation
		color_rgb_to_yuv(img_yuv, img_unsharp, height, width, bit_channel);
	}
	else {
		mono_to_yuv(img_yuv, img_unsharp, height, width, bit_channel);
	}
	
	/*
	 * In case of local sharpening, do calculate sharpening mask.
	 */
	if(sharp_alg == SHARP_ALG_LOCAL) {
		filter_sobel_3s16(img_sobel, img_yuv, height, width, 0, 1, 1);
		filter_gauss_3s16(img_gauss, img_sobel, height, width, 3, 1.0, 0, 1, 1);		// incresing the kernel size need more computing performance
		local_sens = (int)((1.0-local_sens/100.0)*(1<<bit_channel));
		filter_binary_3s16(sharp_mask, img_gauss, height, width, local_sens, (1<<bit_channel)-1, 0, 1, 1);
	}
	
	/*
	 * Y-channel is sharpened only to avoid color shifting
	 */
	make_sharper(img_yuv_sharp, img_yuv, height, width, sharp_factor, (1<<bit_channel)-1, sharp_alg == SHARP_ALG_LOCAL ? 1:0, sharp_mask);
	
	// YUV to RGB transformation
	if(is_color) {
		color_yuv_to_rgb(img_sharp, img_yuv_sharp, height, width, bit_channel);
	}
	else {
		yuv_to_mono(img_sharp, img_yuv_sharp, height, width, bit_channel);
	}
	return 0;
}