Structure-Preserving Fractal Image Generation Using the Proposed Image Morphology-Based Retargeting Algorithm

Abstract

To construct fractal images, affine transformations are iteratively applied to an initial set of images. In the existing affine transformation framework, a scaling operator is generally used to change the aspect ratio of an image. However, the incorporation of the scaling operator may produce structural deformation in salient regions of the image. As a result, prominent reflective regions may become distorted as the number of iterations during fractal image generation increases significantly. To minimize deformation in reflective regions, this paper proposes an efficient affine transformation that incorporates a Proposed Image Morphology-Based Retargeting Algorithm (PIMRA) to replace the conventional scaling operator. To achieve this objective, the entire system is divided into two models. In the first model, algorithms for translation, rotation, and shearing are proposed. To modify the image dimensions in a content-aware manner, PIMRA is utilized. To evaluate the effectiveness of all the proposed algorithms, parameters such as luminance , contrast , and structure  are considered. Objective Image Quality Assessment (OIQA) is performed using the Structural Similarity Index Measure (SSIM). The IQA results show that the performance of PIMRA is 3.25% higher than that of the existing scaling operator. In the second system model, all the proposed geometric transformation techniques are combined to develop an affine transformation framework. The obtained values from IQA showcase that the performance of PAT is 10.45% and 12.45% higher than TAT when the images are selected from the RetargetMe and SIR² dataset respectively.