CODEWITHSUNDEEP
ccuda
dangerChihuahua007
dangerChihuahua007

Reputation: 20895

How do I read the RGB channels of an image in C?

I am using pycuda to compute the intensity of the pixels of an image. To this end, I have sent the image to the GPU as follows.

 img = np.float32(np.array(img.imread('my_image.jpg')))
 img_gpu = gpuarray.to_gpu(img)

Then, in a kernel (written in c of course), I want to get the RGB values as follows (in pseudocode).

__global__ void get_intensities(float* img, float* intensities) {
    intensities[globalIndex] = R(x, y) + G(x, y) + B(x, y)
}

My big problem now is getting the RGB channels in C. How do I do that?

Upvotes: 1

Views: 1264

Answers (1)

wsun
wsun

Reputation: 193

Though this ultimately depends on how the image is stored, I would take advantage of structs in this situation. For example:

image = np.uint8(np.array(img.imread('my_image.jpg')))
img_gpu = gpuarray.to_gpu(image)

__global__ void intensity(uchar3* img, int* intensities)
{
    ...
    intensities[globalIndex] = img[globalIndex].x + img[globalIndex].y + img[globalIndex].z;
    ...
}

See http://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#built-in-vector-types

Upvotes: 1

Related Questions