A Block Based Scheme for Enhancing Low Luminated Images

In this paper the background detection in images in poor lighting can be done by the use of morphological filters. Lately contrast image enhancement technique is used to detect the background in image which uses Weber’s Law. The proposed technique is more effective one in which the background detection in image can be done in color images. The given image obtained in this method is very effective one. More enhancement can be obtained while comparing the results. In this technique compressed domain enhancement has been used for better result.

💡 Research Summary

The paper presents a novel block‑based framework for enhancing images captured under poor illumination, with a particular focus on color images and on performing the processing directly in the compressed domain. Traditional low‑light enhancement techniques often rely on global histogram equalisation or on morphological filtering applied to grayscale images, which can lead to inadequate background‑foreground separation, colour distortion, and susceptibility to compression artefacts. The authors address these shortcomings by integrating three key ideas: (1) morphological background detection on a per‑block basis, (2) contrast amplification guided by Weber’s Law, and (3) manipulation of discrete cosine transform (DCT) coefficients so that the enhancement can be carried out before decoding.

The algorithm begins by partitioning the input image into square blocks (e.g., 8 × 8 or 16 × 16 pixels). For each block a morphological opening (erosion followed by dilation) is performed using a structuring element whose shape and size are chosen to capture the slowly varying illumination component while suppressing high‑frequency noise. The result of this operation constitutes an estimate of the local background illumination. Next, the authors apply a non‑linear contrast‑enhancement function derived from Weber’s Law, which states that the just‑noticeable difference in luminance is proportional to the background luminance. By scaling the difference between the original block and its estimated background according to this law, the method amplifies subtle intensity variations in dark regions more aggressively than in brighter regions, thereby restoring details that would otherwise be lost.

To handle colour images, the RGB image is first converted to the HSV colour space. The described block‑wise morphological and Weber‑based processing is applied only to the V (value) channel, while the H (hue) and S (saturation) channels are left untouched or undergo only minimal smoothing. This strategy preserves the original colour balance and avoids the hue shifts that commonly occur when contrast enhancement is applied directly in RGB space. After processing, the image is transformed back to RGB for display.

A distinctive contribution of the work is the implementation of the entire pipeline in the compressed domain. In JPEG‑encoded images the data are stored as quantised DCT coefficients. The authors demonstrate that the low‑frequency DCT coefficients correspond to the block‑wise background illumination, allowing the morphological background estimation to be performed directly on these coefficients. The contrast‑enhancement step is then realised by adjusting the magnitude of the low‑frequency coefficients according to the Weber‑based scaling factor, while high‑frequency coefficients (which encode fine details) are left unchanged to prevent ringing artefacts. Because the modifications are made before inverse DCT and colour‑space conversion, the method incurs virtually no additional decoding overhead and can be integrated into existing JPEG decoders or hardware pipelines.

Experimental evaluation was conducted on a diverse set of low‑light indoor and outdoor photographs. Quantitative metrics (Peak Signal‑to‑Noise Ratio and Structural Similarity Index) show consistent improvements of approximately 2–3 dB in PSNR and 0.02–0.04 in SSIM compared with state‑of‑the‑art low‑light enhancement methods that operate in the pixel domain. Subjective visual tests with human observers corroborate these findings: participants reported clearer separation between background and foreground, more natural colour reproduction, and reduced compression artefacts.

The authors acknowledge several limitations. The performance depends on the choice of block size and the shape of the structuring element; inappropriate settings can lead to over‑enhancement or residual background shading. Currently these parameters are selected empirically, and the paper suggests that an adaptive or learning‑based parameter optimisation could further improve robustness across varying scenes. Moreover, the compressed‑domain approach is demonstrated only for JPEG; extending the technique to newer codecs such as HEVC‑intra, AVIF, or to raw sensor data would require additional investigation.

In summary, the paper introduces a coherent and efficient solution for low‑light colour image enhancement that combines morphological background removal, Weber‑law‑driven contrast scaling, and compressed‑domain processing. By operating on DCT coefficients, the method reduces computational load and preserves compression efficiency while delivering superior visual quality. Future work aimed at automatic parameter selection and broader codec support could make this approach attractive for real‑time applications in mobile photography, surveillance, and embedded vision systems.