In the context of image coding systems, the fractal image encoding and decoding method is a better option due to its capacity for high achievement in compression ratio and resolution independence at any scale. Fractal image coding embodies two major phases. Firstly, the encoding step, where we produce numerical data, called Iterated Function System (IFS) data, from the statistically self-similar input image. Secondly, in the decoding step, we decode IFS data to revert to a fractal image, as an attractor, by using an inverse function. Notwithstanding, the popularity of the fractal encoder in light of the partition iterative function system has deteriorated into nearly out of favour due to its massive encoding time. So, the existing process is still not applicable. This paper proposes a new approach in which block dimensions are adjusted to address the limitation of requiring a high encoding time. The new strategy is designed for efficient pooling selections. It uses the central pixel and its neighbors based on gradient values, giving more weight to pixels with higher gradient magnitudes. This allows the matching process to take into account not just the pixel intensity but the change in intensity across the block, making it more robust to noise or slight variations in pixel values. This study confirms through experiments acceptable improvement in both encoding time and quality of the decoded image. Subsequently, it is considered an enhancement to some of the previous methods in the literature.