For the planning and development of rural and urban areas, land cover segmentation of plant and water body areas from satellite images is essential. Large sample image datasets are needed for train the land cover categorization algorithms that are currently in use. Due to heterogeneous pixel as well as geometric distortion over boundaries and curvature zone, classifying the land cover of plant and water areas using existing algorithms is a difficult process. The accurate categorization and measuring of land cover are impacted by mixed pixels. Geometric distortion affects the contour of the land cover and results from the framing of isotropic and angular selectivity while image acquisition. Without training datasets, the intent of Transverse dyadic wavelet transform (TDyWT)[2] in this study improves and classifies the vegetation and water area in the land cover of LANDSAT images. The proposed TDyWT employs Cantor 5x9x7 wavelet for reconstructing and Haar wavelet for decomposition. Due to the wavelet's reversible and lifting qualities, the TDyWT improves the contour, curvature, and boundary[1] of plant and water areas in LANDSAT images. Geometric distortion and spatial scale[1] mistake of mixed pixels are eliminated by TDyWT. By total station and errors modelling methodologies, spatial scale inaccuracy in classic land surveying is eliminated. According to the findings, compared to ground truth survey methods, the introduced TDyWT algorithm identifies the area of subclasses of vegetative and water with 95% accuracy.The proposed approach was evaluated on a dataset of satellite images and achieved high accuracy in measuring water body areas. The results demonstrate the potential of deep learning methods for accurately and efficiently monitoring water resources from satellite imagery.