People are at a much higher risk now because of the new coronavirus that causes COVID-19 disease. The virus is quickly spreading evetywhere in the world. Therefore, it is crucial that we create rapid diagnostic tools to identify infected people and stop the virus from spreading. The development of ML would allow us for early diagnosis of COVID-19, allowing for preventive measures to be taken as soon as feasible. But because of small sample numbers, especially in chest X-rays, it is harder to find this condition. Recent work has developed CNN approaches that can recognise COVID-19 from X-ray images. In the real world, it's preferable to acquire the input of many medical professionals before making a major healthcare decision. The reliability of a medical diagnosis is improved when doctors reach a consensus. Since all of the prevalent CNN models are conceptually similar, we advocated for their combined usage. They have learned to infer things on their own. The models are then integrated using a cutting-edge technique called weighted average ensembling to forecast a class value. The new method of assembly should result in a more precise estimate. Our strategy utilises three pre-trained CNN models: DenseNet201, Resnet50V2, and VGG-19. We have also used the Discrete Wavelet Transform (DWT) method for denoising, U-Net for Image segmentation, SMOTE for data balancing, Keras Tuner for hyperparameter optimisation for better accuracy in result when applied to X-ray pictures, our suggested model successfully have achieved 99.17% of accuracy while detecting COVID-19.