Trade-off between Image Quality and Computational Complexity: Image Resizing Perspective

Abstract

This study proposed a new approach for resizing image deal with quality and computational complexity. Here, previous methods in image resizing do analytical works to approximate the original picture element (pixel) or to remove high frequency coefficients. For images with huge pixel, this will result in computational burden due to number of multiplication and addition in the synthesized formula. Instead of the works, this study proposed a new approach in removing the coefficients by exploiting the second-order block matrix without the need to synthesize the formula. It can be called a fully numeric image resizing method. The result shows that the resized version of original image has peak signal to noise ratio (PSNR) equal to 35.24 dB for resizing the famous Lena image which means compareable to the conventional which has PSNR value around 35 dB but here deriving analytical formula is not required. Reducing computational complexity is also achieved as expected with result only 16 addition involved with no multiplication required. This is lower than the conventional in term of computational complexity. Overall, the proposed method has a good balance for both performances than the conventional approaches.

Keywords

• computational complexity
• filtering formula
• image quality
• image resizing
• resizing matrix