C - Bilinear scaling (interpolation) byte by byte [closed]

Question

Closed. This question needs to be more focused. It is not currently accepting answers.

---

Want to improve this question? Update the question so it focuses on one problem only by editing this post.

Closed 3 years ago.

Improve this question

UPDATE: Code is OK now, see edits at the end of question

I'm writing a simple application that is supposed to scale given image and display the result on the screen. Image loading, displaying etc is achieved throught SDL, but I still have problem with scaling function - it produces gibberish.

I have to operate on 24-bit images, thus the uint8_t casting and byte by byte calculations.

#include <stdint.h>
void blin(uint8_t* pixelsIn, uint8_t* pixelsOut, int w, int h, float scale)
{
    int index1, index2;
    int w2, h2;
    int i, j, k;
    float x, y;
    float t;
    int p1, p2;

    w2 = (int)(scale*w + 0.5);
    h2 = (int)(scale*h + 0.5);
    p1 = w*3;
    if(p1%4) p1 += (4-p1%4);
    p2 = w2*3;
    if(p2%4) p2 += (4-p2%4);
    for(i=0;i<h2;i++) //line
    {
        index2=i*p2;
        for(j=0;j<w2;j++) //column
        {
            x=((float)(j))/scale;
            index1=(int)(x) * 3;
            x-=(int)(x);
            y=((float)(i))/scale;
            index1+=(int)(y) * p1;
            y-=(int)(y);
            for(k=0;k<3;k++) //for color in R, G, B
            {
                t = (float)(pixelsIn[index1]) * (1.0-x)*(1.0-y);
                t += (float)(pixelsIn[index1+3]) * (x)*(1.0-y);
                t += (float)(pixelsIn[index1+p1]) * (1.0-x)*(y);
                t += (float)(pixelsIn[index1+p1+3]) * (x)*(y);
                pixelsOut[index2] = (uint8_t)(t);
                index1++;
                index2++;
            }
        }
    }
}

Edit: obvious error, index2 was not zeroed and x was calculated without multiplying by 3 (bytes per pixel). But image is still not properly scaled, this is before and after for scale=1.0 (jpgs just for faster upload):

• Before: http://i.stack.imgur.com/TZLF4.jpg
• After: http://i.stack.imgur.com/IWKfh.jpg

Edit2: 2nd problem was 4-byte alignment inside SDL_Surface pixel structure. Now it works like a charm (code here is updated), though it's intended to work only on 24bit images - see comments to best answer.

Answer 1

I think this line:-

index1=(int)(x);

should be:-

index1=(int)(x)*3;

Also, don't assume that the stride is the same as width.sizeof(pixel), the first pixel on a line may be aligned to a word/dword/etc boundary, so I'd change the code to:-

void blin(uint8_t* pixelsIn, uint8_t* pixelsOut, int w, int h, float scale, int input_stride, int output_stride)
{
    int index1, index2;
    int w2, h2;
    int i, j, k;
    float x, y;
    float t;

    w2=(int)(scale*w + 0.5);
    h2=(int)(scale*h + 0.5);
    index2=0;
    for(i=0;i<h2;i++) //line
    {
        int pixelindex2=index2;
        for(j=0;j<w2;j++) //column
        {
            x=((float)(j))/scale;
            index1=(int)(x)*3;
            x-=(int)(x);
            y=((float)(i))/scale;
            index1+=(int)(y) * input_stride;
            y-=(int)(y);
            for(k=0;k<3;k++) //for color in R, G, B
            {
                t = (float)(pixelsIn[index1]) * (1.0-x)*(1.0-y);
                t += (float)(pixelsIn[index1+3]) * (x)*(1.0-y);
                t += (float)(pixelsIn[index1+w*3]) * (1.0-x)*(y);
                t += (float)(pixelsIn[index1+w*3+3]) * (x)*(y);
                pixelsOut[pixelindex2] = (uint8_t)(t);
                index1++;
                pixelindex2++;
            }
        } //column
        index2+=output_stride;
    } //line
}

where input_stride and output_stride are the number of bytes between the start of consecutive lines which may not be the same as width * 3.

You may also want to consider making the '3' constant a variable so that you can deal with different image formats.