• home
• current students
• graduate
• events

Presenter: Ali Mostafavian
Supervisor:

Date: Thu, July 26, 2018
Time: 13:00:00 - 00:00:00
Place: EOW 430

ABSTRACT

bstract

In this seminar, I will present  the highlights of my PhD thesis. Triangle-mesh models, as one of the approaches for representing images based on nonuniform sampling, have become quite popular and beneficial in many applications. In this thesis, image representation using triangle-mesh models and its application in image scaling are studied. In particular, two new mesh-generation methods, namely, the SEMMG and MISMG methods are proposed to produce mesh models that are effective for image representation and image scaling, respectively. Moreover, the MISMG scheme is combined with a subdivision-based model-rasterization algorithm to yield a novel edge-preserving mesh-based technique, called the MIS method, for image scaling.

Evaluation results show that the reconstructed images obtained from the SEMMG method are better than those obtained by the competing methods in terms of both PSNR and subjective quality. More specifically, we show that the PSNR of the reconstructed images produced by the SEMMG method are on average 5.37, 4.07, 3.79, 3.44, 1.83, and 0.9 dB higher than those obtained by the ED, MGH, GVS, ERDGPI, BSP, and HWT methods, respectively. Furthermore, for a given PSNR, the SEMMG method is shown to produce much smaller meshes compared to those obtained by the GVS and BSP methods, with approximately 60-80\% fewer vertices and 10-60\% fewer triangles, respectively. 

Besides the superior image approximations achieved with the SEMMG method, this work also makes contributions by studying the application of triangle-mesh mesh models in image scaling, leading to the proposal of the novel MIS method for image scaling. Results of the subjective comparison with four well-known scaling methods show that the proposed MIS method outperforms all of them by producing scaled images of better quality with more accurate and sharper edges. Moreover, the well-known percentage edge error (PEE) metric is also used to measure the sharpness of the reconstructed edges in the scaled images produced by the methods under comparison. As a result, at a scaling factor of 4, the PEE results show that the scaled images obtained by our MIS method are approximately 22\%, 16\%, 18\%, and 16\% sharper than those obtained from the bilinear, bicubic, DCCI, and NEDI interpolation methods, respectively. These significant reductions in blurring artifacts yield scaled images with sharper edges that look more pleasant to a human observer.