The objective of camouflaged object detection (COD) is to identify objects that highly resemble the background. Existing methods for COD mainly explore features in the RGB space. However, unobvious differences between camouflaged objects and backgrounds are difficultly captured by using the spacial convolution operations in the colour space, which causes missed detection in object levels and boundary regions. Inspired by that frequency-domain features in certain wavelet frequency bands may better represent the visually unobvious differences, we proposes a Wavelet Transform based Feature Enhancement network (WTFENet) for COD. Firstly, the WTFE module is used to extract contour and context information for coarsely locating the camouflaged objects. Further, to refine the coarse localization and establish a more precise boundary at the level of individual pixels, a context feature fusion module (CFM) is designed to exchange learned information between frequency domain and colour space by integrating the details from both specific high-frequency wavelet bands and spacial texture information. Moreover, an F-Loss is designed to drive the fusion procedure for better addressing the pixel-level misdetection along the object boundary. Compared to existing methods, our proposed approach demonstrates competitive qualitative and quantitative performances on three commonly-used benchmark datasets.