Beam shaping is a crucial laser technology for achieving arbitrary light intensity distributions, typically achieved through Diffractive Optical Element (DOE). The Gerchberg-Saxton (GS) algorithm is an efficient method for DOE design, while problems such as local optima and poor uniformity remain unsolved. Therefore, improving the GS algorithm has become a research hotspot in diffractive optics. To increase diffraction efficiency and effectively suppress speckle noise, an improved GS algorithm based on feedback from non-signal region speckle is proposed in this paper. The core idea is as follows: first, the initial phase distribution is generated by using Zernike polynomial combined with polynomial weight coefficient Z. The weight coefficient Z is optimized through global amplitude constraint to determine the optimal Z value within a specific range, thereby resulting in an optimal initial phase form. Then, a basic target spot with high diffraction efficiency is obtained by GS algorithm, and the speckle feedback algorithm in non-signal region is used to suppress the speckle noise in non-signal region and improve the uniformity of signal region. In order to verify the effectiveness and generalization capability of the improved algorithm, the optimization designs were carried out for triangular spot, circular spot, and square spot, and the optical simulations and optical experiments were conducted. When shaping the Gaussian beam into a triangular spot under the constraint of RMSE<0.1%, the simulation results show that the diffraction efficiency of the improved algorithm reaches 98.30%, significantly higher than the87.49% achieved by the MRAF algorithm. The Root Mean Square Error (RMSE) is 0.27%, which is much lower than the 32.78% achieved by the GS algorithm. In addition, for circular and square spots, the results of the improved algorithm are also superior to the other two algorithms, demonstrating strong generalization ability for different spot shapes. The experimental results show that the improved algorithm has a speckle contrast of 0.020 9, while the GS algorithm has a speckle contrast of 0.280 1, which indicates that the improved algorithm has higher internal uniformity generation ability. Meanwhile, the Peak Background Ratio (PBR) is used to evaluate the speckle noise in the non-signal region. The PBR of the improved algorithm is 2.984 1, which is significantly higher than the 0.008 1 achieved by the MRAF algorithm, indicating that the improved algorithm has almost no speckle noise in the non-signal region, while the MRAF algorithm exhibits severe speckle. At the same time, this paper also discusses the influence of two key factors on the quality of the output light spot: 1) The influence of the waist radius on the output light spot. When the beam waist radius decreases to 600 μm, the overall brightness and uniformity of the spot are reduced, and pixel vortices are generated. However, when the waist diameter is expanded to 1 500 μm, the light spot generated by the improved algorithm can still maintain high diffraction efficiency and uniformity, which indicates that the improved GS algorithm exhibits certain robustness to different Gaussian beam waist radii. 2) The matching relationship between the beam waist and the position of DOE. Deviations from the optimal position significantly affect the output spot quality, leading to a decline in the quality of the spot. Three horizontal deviations of -0.32 mm, 0 mm and 0.32 mm are investigated in this paper, and the results show that these deviations can lead to a decrease in spot uniformity. In summary, when the Gaussian beam is shaped into the target spot, the improved algorithm not only improves the internal uniformity of the target spot, but also effectively inhibits the speckle noise in the non-signal region, while maintaining a high diffraction efficiency. The effectiveness of the improved algorithm is verified by optical experiments and simulations. The method proposed in this paper provides a valuable reference for designing DOE with high diffraction efficiency and uniformity.