Abstract
MATLAB is one of most commonly used platforms in multiple scientific applications like digital image processing, digital signal processing etc.
The high level programming syntax and friendly user of MATLAB makes it best suited to write technical code. Among, the significance of MATLAB's programming is its rich library of built-in function that makes programming more easily. Nerveless, the standard MATLAB uses an interpreter which decelerates the processing, particularly while executing loops. This becomes a bottleneck performance in programs that make excessive use of loops. Therefore, MATLAB is frequently exposed to the memory latency and the issues of slow execution. In order to accelerate MATLAB's processing, we use NIVIDIA'S CUDA parallel processing architecture. We note that processing can be speed up significantly by interfacing MATLAB with CUDA and parallelizing the most time consuming portion of MATLAB's code white balance. The obtained results indicate, that the speedup is proportional to the image size until it attains a maximum at 2056*3088 pixels, beyond these values the speedup decreases. The performance with GPU enhances above a factor of 14 similar to 15 compared with CPU.