In ultrasonic tests,the redundant signal and random noise are easily to be generated when the sensor collects data.Compressing and reconstructing the signal or noise always get the researchers’ attention.Based on the theory of compressive sensing and ultrasonic testing method,an improved CoSaMP for tube fouling ultrasonic testing compression and reconstruction was proposed:The K-SVD algorithm was used to sparsely represent the ultrasonic testing signal.The relationship between the peak signal to noise ratio and the number of iterations was derived to determine the sparse degree adaptively.Finally,the sparsity differences of the echo signal and random noise in the transform domain was used to pick up effective information.The results showed that the improved CoSaMP algorithm can accurately locate signal sparsity.The reconstructed signal reflects that the characteristics of the original signal and the operation time are shorter.That proves the feasibility of the improved CoSaMP algorithm.