Improving low light images is challenging because of the lack of proper lighting, and it becomes even harder when there is no reference image to guide the enhancement process. This challenge has led to an increased interest in low-light image enhancement (LLIE) methods. Many existing approaches are based on reference images, making them less effective in real-world situations like surveillance and autonomous driving, where capturing paired low-light images is often impractical. Additionally, most non-reference LLIE methods process all three RGB (Red, Green, Blue) channels while optimizing for illumination smoothness, spatial consistency, exposure control, and color balance simultaneously. This multi-channel approach, combined with non-reference loss functions, increases computational complexity and reduces the effectiveness of models. To address these challenges, this work proposes a novel non-reference LLIE method that focuses on single-channel processing with targeted loss functions. Our approach operates in the HSV (Hue, Saturation, Value) color space, enhancing only the V channel. This channel-specific approach enables precise pixel-wise brightness adjustments without the need for reference images or multi-channel processing. This specialized deep network is trained to predict pixel-wise curves for dynamic range adjustment using two carefully chosen non-reference loss functions: exposure loss and illumination loss. This targeted approach ensures effective enhancement while maintaining computational efficiency. The proposed method is evaluated using Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), Mean Absolute Error (MAE), Naturalness Image Quality Evaluator (NIQE), and Learned Perceptual Image Patch Similarity (LPIPS), demonstrating its effectiveness in enhancing low-light images.