Watermark detection software is obfuscated using a table to hide embedding and detection algorithms. As the table size is limited, the block size is also limited for watermarking. To address this situation, a new quasi-chirp transform is developed to improve embedding efficiency. The quasi-chirp transform is different from the conventional DCT or Fourier transform. It contains multiple frequency components in a single basis of the transform. It disperses image data rather than compressing it, as the DCT does. The dispersed data increases the range for embedding watermarks. The chirp transform is able to embed even on a flat area of an image. Using this chirp transform, embedding and detection experiments for image data with small block sizes were carried out. A high SNR and robust watermark with an evaluated obfuscation were obtained.