FPGA实现YCbCr444转RGB

1 基本概念

颜色空间（color space）是颜色集合的数学表示。三个最常用的颜色模型是：RGB（用于计算机图像学中）；YIQ、YUV 或 YCbCr（用于视频系统中）；CMYK(用于彩色打印)。

1.1 YCbCr 颜色空间

YCbCr 颜色空间是 YUV 颜色空间的缩放和偏移版本。Y 定义为 8bit,标称颜色范围为 16-235；Cb 和 Cr 标称颜色表示范围为 16-240。YCbCr 的采样格式一般有 4:4:4、4:2:2、4:1:1、和 4:2:0。

4:4:4 YCbCr 格式：

图 1 表示 4:4:4 格式 YCbCr 采样点的定位。每个采样点有 Y、Cb 和 Cr 值，每个颜色值的颜色分量为 8bit（典型），因此每个采样点 24bit。

图 1 4:4:4 采样

1.2 RGB 颜色空间

红、绿和蓝（RGB）颜色空间广泛用于计算机图像学和显示器。红绿蓝是三种基本的加性颜色，可以用三维的笛卡尔坐标系统来表示 RGB 颜色空间。

1.3 YCbCr 到 RGB 颜色空间的转换：数学公式

为了将标称取值范围为 16~235（Studio R’G’B’）的 8 位 YCbCr 数据转换为 R’G’B’颜色公式可以简化为：

R’=Y+1.371(Cr-128)

G’=Y-0.689(Cr-128)-0.336(Cb-128)

B’=Y+1.732(Cb-128)

2 matlab 实现 ycbcr444 转 RGB

close all
clear all
clc
 
I=imread('1.bmp');
 
[H ,W ,D]=size(I);
 
R=double(I(:,:,1));
G=double(I(:,:,2));
B=double(I(:,:,3));
 
 
Y0= double(zeros(H,W));
Cb0 =double(zeros(H,W));
Cr0 = double(zeros(H,W));
 
R0= double(zeros(H,W));
G0 =double(zeros(H,W));
B0 = double(zeros(H,W));
 
%RGB 转 YCbCr444
 for i = 1:H
     for j = 1:W
         Y0(i, j) = 0.299*R(i, j) + 0.587*G(i, j) + 0.114*B(i, j);
         Cb0(i, j) = -0.172*R(i, j) - 0.339*G(i, j) + 0.511*B(i, j) + 128;
         Cr0(i, j) = 0.511*R(i, j) - 0.428*G(i, j) - 0.083*B(i, j) + 128;
     end
 end 
 
 for i = 1:H
     for j = 1:W
         RGB(i, j,1) =Y0(i,j)+1.371*(Cr0(i,j)-128);
         RGB(i, j,2) =Y0(i,j)-0.689*(Cr0(i,j)-128)-0.336*(Cb0(i,j)-128);
         RGB(i, j,3) =Y0(i,j)+1.732*(Cb0(i,j)-128);
     end
 end
 
YCbCr(:,:,1)=Y0;
YCbCr(:,:,2)=Cb0;
YCbCr(:,:,3)=Cr0;
 
YCbCr=uint8(YCbCr);
 
RGB=uint8(RGB);
 
figure(1),
imshow(YCbCr),title('YCbCr');
 
figure(2),
imshow(RGB),title('RGB');

转之前（YCbCr444）

   转之后（RGB）

3 fpga 实现

/*
	计算公式：	
	R = 1.164(Y - 16) + 1.793(CR - 128)                   = 1.164Y           + 1.793CR - 248.128;
	G = 1.164(Y - 16) - 0.534(CR - 128) - 0.213(CB - 128) = 1.164Y - 0.213CB - 0.534CR + 76.992;
	B = 1.164(Y - 16) 		    + 2.115(CB - 128) = 1.164Y + 2.115CB           - 289.344;
	其中，时序在计算过程中完全没有用到
	输入到输出有三个 clock 的时延。
	第一级流水线计算所有乘法；
	第二级流水线计算所有加法，把正的和负的分开进行加法；
	第三级流水线计算最终的和，若为负数取 0；
*/
`timescale 1ns/1ps
module	ycbcr_to_rgb(
	input	clk,
	input	wire[7 : 0]		i_y_8b,
	input	wire[7 : 0]		i_cb_8b,
	input	wire[7 : 0]		i_cr_8b,
 
	input	                i_h_sync,
	input	                i_v_sync,
	input	                i_data_en,
 
	output	wire[7 : 0]		o_r_8b,
	output	wire[7 : 0]		o_g_8b,
	output	wire[7 : 0]		o_b_8b,
 
	output	reg				o_h_sync,
	output	reg				o_v_sync,                                                                                                    
	output	reg				o_data_en                                                                                                   
);

/***************************************parameters*******************************************/
//multiply 256
parameter	para_1164_10b = 10'd297;	//1.160
parameter	para_1793_10b = 10'd459;	//1.793
parameter	para_0534_10b = 10'd137;	//0.535
parameter	para_0213_10b = 10'd54;		//0.211
parameter	para_2115_10b = 10'd541;	//2.113
parameter	para_248128_18b = 18'd63521;//248.128
parameter	para_76992_18b = 18'd19710; //76.992
parameter	para_289344_18b = 18'd74072;//289.344
/********************************************************************************************/

/***************************************signals**********************************************/
wire				sign_r;
wire				sign_g;
wire				sign_b;
reg[17 : 0]	mult_y_for_r_18b;
reg[17 : 0]	mult_y_for_g_18b;
reg[17 : 0]	mult_y_for_b_18b;

reg[17 : 0]	mult_cb_for_g_18b;
reg[17 : 0]	mult_cb_for_b_18b;

reg[17 : 0]	mult_cr_for_r_18b;
reg[17 : 0]	mult_cr_for_g_18b;

reg[17 : 0]	add_r_0_18b;
reg[17 : 0]	add_g_0_18b;
reg[17 : 0]	add_b_0_18b;

reg[17 : 0]	add_r_1_18b;
reg[17 : 0]	add_g_1_18b;
reg[17 : 0]	add_b_1_18b;

reg[17 : 0] result_r_18b;
reg[17 : 0]	result_g_18b;
reg[17 : 0]	result_b_18b;

reg			i_h_sync_delay_1;
reg			i_v_sync_delay_1;
reg			i_data_en_delay_1;

reg			i_h_sync_delay_2;
reg			i_v_sync_delay_2;
reg			i_data_en_delay_2;

/********************************************************************************************/

/***************************************initial**********************************************/
initial
begin
	mult_y_for_r_18b <= 18'd0;
	mult_y_for_g_18b <= 18'd0;
	mult_y_for_b_18b <= 18'd0;
 
	mult_cb_for_g_18b <= 18'd0;
	mult_cb_for_b_18b <= 18'd0;
 
	mult_cr_for_r_18b <= 18'd0;
	mult_cr_for_g_18b <= 18'd0;

 
	add_r_0_18b <= 18'd0;
	add_g_0_18b <= 18'd0;
	add_b_0_18b <= 18'd0;
 
	add_r_1_18b <= 18'd0;
	add_g_1_18b <= 18'd0;
	add_b_1_18b <= 18'd0;
 
	result_r_18b <= 18'd0;
	result_g_18b <= 18'd0;
	result_b_18b <= 18'd0;
 
	i_h_sync_delay_1 <= 1'd0;
	i_v_sync_delay_1 <= 1'd0;
	i_data_en_delay_1 <= 1'd0;
 
	i_h_sync_delay_2 <= 1'd0;
	i_v_sync_delay_2 <= 1'd0;
	i_data_en_delay_2 <= 1'd0;

 
	o_h_sync <= 1'd0;
	o_v_sync <= 1'd0;                                                                                                  
	o_data_en <= 1'd0;                                             	
end 
/********************************************************************************************/
 
/***************************************arithmetic*******************************************/
//LV1 pipeline : mult
always @ (posedge	clk)
begin
	mult_y_for_r_18b <= i_y_8b * para_1164_10b;
	mult_y_for_g_18b <= i_y_8b * para_1164_10b;
	mult_y_for_b_18b <= i_y_8b * para_1164_10b;
end

always @ (posedge	clk)
begin
	mult_cb_for_g_18b <= i_cb_8b * para_0213_10b;
	mult_cb_for_b_18b <= i_cb_8b * para_2115_10b;
end

always @ (posedge	clk)
begin
	mult_cr_for_r_18b <= i_cr_8b * para_1793_10b;
	mult_cr_for_g_18b <= i_cr_8b * para_0534_10b;
end
//LV2 pipeline : add
always @ (posedge	clk)
begin
	add_r_0_18b <= mult_y_for_r_18b + mult_cr_for_r_18b;
	add_r_1_18b <= para_248128_18b;
	add_g_0_18b <= mult_y_for_g_18b + para_76992_18b;
	add_g_1_18b <= mult_cb_for_g_18b + mult_cr_for_g_18b;
	add_b_0_18b <= mult_y_for_b_18b + mult_cb_for_b_18b;
	add_b_1_18b <= para_289344_18b;
end
//LV3 pipeline : y + cb + cr
assign	sign_r = (add_r_0_18b >= add_r_1_18b);
assign	sign_g = (add_g_0_18b >= add_g_1_18b);
assign	sign_b = (add_b_0_18b >= add_b_1_18b);
always @ (posedge	clk)
begin
	result_r_18b = sign_r ? (add_r_0_18b - add_r_1_18b) : 18'd0;
	result_g_18b = sign_g ? (add_g_0_18b - add_g_1_18b) : 18'd0;
	result_b_18b = sign_b ? (add_b_0_18b - add_b_1_18b) : 18'd0;
end

//output 溢出处理
assign	o_r_8b = (result_r_18b[17:16] == 2'b00) ? result_r_18b[15 : 8] : 8'hff;
assign	o_g_8b = (result_g_18b[17:16] == 2'b00) ? result_g_18b[15 : 8] : 8'hff;
assign	o_b_8b = (result_b_18b[17:16] == 2'b00) ? result_b_18b[15 : 8] : 8'hff;
/********************************************************************************************/

/***************************************timing***********************************************/
always @ (posedge	clk)
begin
	i_h_sync_delay_1 <= i_h_sync;
	i_v_sync_delay_1 <= i_v_sync;
	i_data_en_delay_1 <= i_data_en;

	i_h_sync_delay_2 <= i_h_sync_delay_1;
	i_v_sync_delay_2 <= i_v_sync_delay_1;
	i_data_en_delay_2 <= i_data_en_delay_1;

 
	o_h_sync <= i_h_sync_delay_2;
	o_v_sync <= i_v_sync_delay_2;
	o_data_en <= i_data_en_delay_2;
end
/********************************************************************************************/
Endmodule

fpga 实现 YCbCr444 转 RGB 效果和 matlab 一致。