This paper proposes a novel image watermarking scheme using entropy of the host and texture of the watermark images. Entropy is evaluated for the discrete wavelet transform coefficients of the host image. The subband with the highest entropy is selected for embedding the texture of the watermark. Arnold transform is used to texturize the watermark and a texture is selected randomly. The level of security is increased by embedding a texture instead of watermark itself. Various experiments are conducted to analyse the proposed entropy based subband selection, linear weight evaluation, watermark embedding and watermark-host extraction. The experimental results prove that this method effectively extracts the watermark and the host images with better imperceptibility.