When compared to the conventional beamformer (CBF), the Capon beamformer often shows a better target detection performance. However, when the array aperture (or, the degrees of freedoms-DOFs) is limited and the signal-to-noise ratio (SNR) is relatively low, the performance of the Capon beamformer will degrade severely. Unfortunately, for the practical application the target of interest is often submerged in loud interferences and high-level fluctuated noises. Thus, it is very hard for a small-aperture array to find the target of interest. In this paper, to improve the beamforming performance of the small-aperture linear array, we propose an improved Capon beamformer based on Fourier integral method (FIM), which is termed as FIM-Capon for short. FIM-Capon firstly uses the basic theory of FIM to obtain 2N-1 outputs (the number of hydrophones in the array is N), which is actually the Toeplitz averaged covariance matrix calculated from the sampled array covariance matrix. And then, FIM-Capon uses the subarray smoothing and the Capon beamformer to process the 2N-1 outputs of the Toeplitz averaged covariance matrix to produce the final beamforming output. The proposed FIM-Capon method have two advantages; one is making use of the noise suppression ability of FIM, and the other is using the extended DOFs coming from the Toeplitz averaging. Hence, FIM-Capon has a higher input SNR and a larger DOF than the traditional Capon beamformer. As a result, when the aperture is limited and the input SNR is low, FIM-Capon shows a better weak target detection performance than the traditional Capon beamformer. Numerical simulations validate the effectiveness of the proposed FIM-Capon method.