Home // International Journal On Advances in Software, volume 10, numbers 3 and 4, 2017 // View article

A Non-Linear Method to Interpolate Binary Images Using Location and Neighborhood Adaptive Rules

Authors:
Pullat Joy Prabhakaran
Palanganda Ganapathy Poonacha

Keywords: resolution; interpolation; binary-image; thinnes; location-adaptive; neighborhood-adaptiv; corner; slope.

Abstract:
In this paper, we propose a new zooming technique for binary images, using location and neighborhood adaptive, non-linear interpolation rules. These rules are inspired by the way an artist would draw an enlarged image. Using simple examples, we show that the output generated by popular interpolation techniques is very different from what a human does. Our rules are based on such observations for simple objects, and they try to mimic what an artist does. Using these rules, we interpolate complex images and demonstrate the impact. We compare our method with bicubic interpolation and show that our method gives better visual quality. We also show that, on an average, our method results in higher PSNR and a lower MPSNR. The SSIM of the output images are nearly the same for both methods. Our method overcomes a number of problems associated with known interpolation techniques, such as blurring and thickening of edges. Our method uses a set of sixteen rules in five categories. Each pixel in the interpolated image is computed by a chosen rule. The choice depends on the location of the pixel and the content in the neighborhood. The size of the neighborhood varies. Some rules can be influenced by distant pixels in the input and can impact distant pixels in the output. We present examples showing the effectiveness of our method. The results are visually appealing. We show that lines and dots, with single pixel thickness, retain their thickness. Inclined lines and solids don’t develop as much jaggedness as happens with bicubic interpolation. Similarly, curves are also relatively smoother.

Pages: 526 to 538

Copyright: Copyright (c) to authors, 2017. Used with permission.

Publication date: December 31, 2017

Published in: journal

ISSN: 1942-2628