dim3 threadsPerBlock(16, 16);
dim3 blocksPerGrid((width + 15) / 16, (height + 15) / 16);
imageKernel<<<blocksPerGrid, threadsPerBlock>>>(d_image, width, height);
