This paper proposes a novel learnable color-texture weighted CamShift (LCTW-CamShift) algorithm to tackle the problem of small target tracking in UAV videos. The LCTW-CamShift employs a pixel-wise adaptive weighting strategy to modulate each pixel based on chromatic and structural cues. Specifically, features from the hue, saturation, and value color space and rotation-invariant Local Binary Patterns are fused via a lightweight two-layer neural network to generate spatially adaptive weights, which guide the refinement of the back-projection map. The LCTW-CamShift tracker can accurately focus on salient foreground regions while suppressing background clutter and noise. Experiments on the UAV123 benchmark show that LCTW-CamShift consistently outperforms state-of-the-art trackers in both accuracy and robustness, particularly under conditions of color ambiguity, low-texture scenes, and dense backgrounds.