Solution Proposal for Accelerating Interpolation Algorithms in Image Scaling: A Comparative Application

Image processing forms the foundation of computer vision and is widely used in many applications. In scaling operations, images are resized by performing computations on the values of pixel points. The success of computer vision processes is directly related to the use of accurate and high-quality images. Ensuring that source images have the same dimensions is of great importance for effective processing. The use of source images with different sizes adversely affects the performance of computer vision algorithms. Therefore, image scaling operations must be performed both accurately and efficiently. In this study, processing time values are evaluated and compared using Bicubic and Lanczos interpolation-based image scaling methods under different scaling factors (0.5× and 2×) and various parallelization models. Using the developed multithreaded console application, thread configurations are compared under single-core and multi-core execution scenarios. Experimental results have shown that a speed performance increase of up to 176.4% can be achieved with the multi-core parallel model. In addition, when all the results were evaluated, it was concluded that the Bicubic interpolation method was faster than the Lanczos method.