In　this　paper,　we　propose　a　new　type　of　continuous　wavelet　transform.　However　we　discretize　variables　of　integral　a　and　b,　any　numerical　integral　has　a　high　resolution,　and　a　does　not　appear　in　the　denominator　of　the　integrand.　Furthermore,　we　give　two　discrezitation　methods　of　the　new　wavelet　transform.　For　the　one-dimensional　situation,　we　give　quadrature　formula　of　the　discretized　inverse　wavelet　transform.　For　the　multi-dimensional　situation,　we　develop　the　commonly　wavelet　network　based　on　the　discretized　inverse　wavelet　transform　of　the　new　wavelet　transform.　Finally,　the　numerical　examples　show　that　the　continuous　wavelet　transform　constructed　in　this　paper　has　higher　computing　accuracy　compared　with　the　classical　continuous　wavelet　transform.