原创 将CMOS摄像头中的Bayer数据转换为RGB格式

[Linux-uvc-devel] C function to convert from Bayer to RGB24 pixel formats

Vojtech Pavlik vojtech at suse.cz
Thu Jan 11 23:44:47 CET 2007

• Previous message: [Linux-uvc-devel] C function to convert from Bayer to RGB24 pixel formats
• Next message: [Linux-uvc-devel] QuickCam Fusion lockup
• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On Thu, Jan 11, 2007 at 10:47:10PM +0100, Laurent Pinchart wrote:

> Hi Luca,

>

> > This is function in C to convert from Bayer to RGB24 pixel formats.

> > The linear interpolation is more efficient than the ones I saw in other

> > similar functions on internet, as the code omits "if-then-else" conditional

> > statements in the loops.

>

> Thanks for the code. This will be usefull for all the users who are interested

> in raw image processing. There seem to be quite a lot of them lately :-)

That is easy to explain: Those interested in raw processing are looking

for a webcam that can do USB 2.0 and is supported on Linux. Not that

much choice. ;)

Btw, my take on Bayer->RGB processing: (attached)

--

Vojtech Pavlik

Director SuSE Labs

-------------- next part --------------

#include <stdio.h>

#include <stdlib.h>

#include <stdint.h>

#include <unistd.h>

#include <sys/mman.h>

#include <sys/types.h>

#include <sys/stat.h>

#include <fcntl.h>

#include <sys/ioctl.h>

#include <string.h>

#include <sys/select.h>

#define u8 uint8_t

#define u16 uint16_t

#define u32 uint32_t

static uint8_t rgb[640*480*3];

#define R(x,y) pRGB24[0 + 3 * ((x) + 640 * (y))]

#define G(x,y) pRGB24[1 + 3 * ((x) + 640 * (y))]

#define B(x,y) pRGB24[2 + 3 * ((x) + 640 * (y))]

#define Bay(x,y) pBay[(x) + 640 * (y)]

static void bayer_copy(u8 *pBay, u8 *pRGB24, int x, int y)

{

        G(x + 0, y + 0) = Bay(x + 0, y + 0);

        G(x + 1, y + 1) = Bay(x + 1, y + 1);

        G(x + 0, y + 1) = G(x + 1, y + 0) = ((u32)Bay(x + 0, y + 0) + (u32)Bay(x + 1, y + 1)) / 2;

        R(x + 0, y + 0) = R(x + 1, y + 0) = R(x + 1, y + 1) = R(x + 0, y + 1) = Bay(x + 0, y + 1);

        B(x + 1, y + 1) = B(x + 0, y + 0) = B(x + 0, y + 1) = B(x + 1, y + 0) = Bay(x + 1, y + 0);

}

static void bayer_bilinear(u8 *pBay, u8 *pRGB24, int x, int y)

{

        R(x + 0, y + 0) = ((u32)Bay(x + 0, y + 1) + (u32)Bay(x + 0, y - 1)) / 2;

        G(x + 0, y + 0) = Bay(x + 0, y + 0);

        B(x + 0, y + 0) = ((u32)Bay(x - 1, y + 0) + (u32)Bay(x + 1, y + 0)) / 2;

        R(x + 0, y + 1) = Bay(x + 0, y + 1);

        G(x + 0, y + 1) = ((u32)Bay(x + 0, y + 0) + (u32)Bay(x + 0, y + 2)

                        + (u32)Bay(x - 1, y + 1) + (u32)Bay(x + 1, y + 1)) / 4;

        B(x + 0, y + 1) = ((u32)Bay(x + 1, y + 0) + (u32)Bay(x - 1, y + 0)

                        + (u32)Bay(x + 1, y + 2) + (u32)Bay(x - 1, y + 2)) / 4;

        R(x + 1, y + 0) = ((u32)Bay(x + 0, y + 1) + (u32)Bay(x + 2, y + 1)

                        + (u32)Bay(x + 0, y - 1) + (u32)Bay(x + 2, y - 1)) / 4;

        G(x + 1, y + 0) = ((u32)Bay(x + 0, y + 0) + (u32)Bay(x + 2, y + 0)

                        + (u32)Bay(x + 1, y - 1) + (u32)Bay(x + 1, y + 1)) / 4;

        B(x + 1, y + 0) = Bay(x + 1, y + 0);

        R(x + 1, y + 1) = ((u32)Bay(x + 0, y + 1) + (u32)Bay(x + 2, y + 1)) / 2;

        G(x + 1, y + 1) = Bay(x + 1, y + 1);

        B(x + 1, y + 1) = ((u32)Bay(x + 1, y + 0) + (u32)Bay(x + 1, y + 2)) / 2;

}

static void bayer_to_rgb24(u8 *pBay, u8 *pRGB24)

{

        int i, j;

        for (i = 0; i < 640; i += 2)

               for (j = 0; j < 480; j += 2)

                       if (i == 0 || j == 0 || i == 640 - 2 || j == 480 - 2)

                               bayer_copy(pBay, pRGB24, i, j);

                       else

                               bayer_bilinear(pBay, pRGB24, i, j);

}

void writepnm(uint8_t *data, int i)

{

        FILE *fd;

        char name[32];

        sprintf(name, "p5-%04d.pnm", i);

        fd = fopen(name, "w");

        fputs("P6\n", fd);

        fputs("640 480\n", fd);

        fputs("255\n", fd);

        fwrite(data, 640*480*3, 1, fd);

        fclose(fd);   

}

int main(void)

{

        int fdi;

        u8 *ibot;

        if ((fdi = open("bayer.raw", O_RDONLY)) < 0) {

               perror("open");

                return -1;

        }

        if ((ibot = mmap(NULL, 640 * 480 + 640*4, PROT_READ, MAP_SHARED, fdi, 0)) == MAP_FAILED) {

               perror("mmap");

               return -1;

        }

        bayer_to_rgb24(ibot, rgb);

        writepnm(rgb, 0);

        close(fdi);

        return 0;

}

DE2_CCD中bayer图像格式转化为30位标准RGB格式

DE2_CCD中bayer图像格式转化为30位标准RGB格式
     CIS mt9m011使用的是bayer颜色格式，如图1所示（黑的地方像素值为0）。奇数行包括green和red颜色的像素，偶数行包括blue和green颜色的像素。奇数列包括green和blue颜色的像素，偶数列包括red和green颜色的像素。

                                                           图1
过程总述：采用行缓冲+流水线的处理方式，将前一级抓取到的行数据（bayer color pattern），实时转换成标准的30位RGB数据并进行适当的下采样，以便于后继的图像处理及显示。

       首先调用一个altshift_taps 的megafunction，altshift_taps 可以作为数据移动寄存器组，经常在图像处理程序中用到。采集进来一系列像素点数据，需要同时对2行像素点同时进行处理时，就可以用这个altshift_taps.共2个taps，每个taps数据宽度为图像每行像素点的个数(1280)。（针对这个megafunction做一下波形仿真就容易理解了）。
             在格式转换时，用到4个量：
mDATA_0:第M行第N列的像素
mDATAd_0：第M行第N-1列的像素
mDATA_1：第M+1行第N列的像素
mDATAd_1：第M+1行第N-1列的像素
4个量组成一个以mDATA_0(位于左上角)为核心的模板：

模板中必定包含一个R、一个B、2个G像素点，那么R作为当前像素mDATA_0的R值，B作为当前像素mDATA_0的B值，取2个G的平均值作为当前像素mDATA_0的G值（通过取G的高10位，舍去最后一位来实现）。对这bayer颜色格式图，任取一个颜色的像素作为核心元素套进模板，这样就可以很好的理解RAW2RGB程序中关于图像格式转化的那部分代码了。